JP5619348B2 - Mold sheet inspection system - Google Patents

Description

  The present invention relates to a defect inspection apparatus for inspecting a defect of a molded sheet such as an optical film such as a polarizing film or a retardation film (particularly, a long optical film wound and stored in a web shape).

  Conventional molded sheet defect inspection devices use a one-dimensional camera called a line sensor, illuminate the molded sheet with a linear light source such as a fluorescent tube, and the longitudinal direction of the molded sheet along the longitudinal direction of the molded sheet. One still image data is acquired by scanning with a one-dimensional camera from the other end to the other end, and a defect of the molded sheet is inspected based on the one still image data. This still image data usually includes a linear light source image. A linear light source image is an image of light emitted from a linear light source and regularly reflected by the molding sheet and reaching the camera when a linear light source and a camera and a reflecting surface are arranged. In the case where a molding sheet is disposed between the linear light source and the camera, it is an image of light emitted from the linear light source and transmitted through the molding sheet to reach the camera. In this defect inspection apparatus, when the width of the formed sheet is wide, a plurality of line sensors are arranged in the width direction so that the entire width direction of the formed sheet can be inspected.

  However, since this conventional defect inspection apparatus inspects defects of a molded sheet based on one still image data (hereinafter simply referred to as “image data”) for the entire molded sheet, The positional relationship between the inspection target pixel and the linear light source image is one predetermined positional relationship. The defect may appear on the image data only when the positional relationship between the pixel to be inspected (target pixel) and the linear light source image is in a specific positional relationship. For example, bubbles that are one type of defect often appear on the image data only when they are at or near the periphery of the linear light source image. Therefore, the defect may not be detected depending on the position. Therefore, the conventional defect inspection apparatus has only a limited defect detection capability.

Then, the applicant of this application has applied for the defect inspection apparatus of the shaping | molding sheet which improved defect identification capability with respect to the said conventional defect inspection apparatus (refer patent document 1). In this defect inspection apparatus, a molding sheet is illuminated with a linear light source such as a fluorescent tube, and moving data (on the molding sheet) using a two-dimensional camera called an area sensor while continuously conveying the molding sheet in a predetermined direction. A plurality of pieces of image data having different imaging positions are acquired, and defects of the molded sheet are inspected based on the moving image data. In this defect inspection apparatus, since it is possible to determine whether there is a defect based on a plurality of pieces of image data in which the positional relationship between the pixel to be inspected and the linear light source image is different, the defect is more reliably detected than the conventional defect inspection apparatus. Can be detected. Therefore, this defect inspection apparatus has improved defect detection capability than the conventional defect inspection apparatus. If this moving image data is used, it is possible to see how the defect moves with respect to the illumination image.
JP 2007-218629 A (released on August 30, 2007)

  However, according to the study of the present inventor, it has been found that the defect inspection apparatus described in Patent Document 1 also has room for improvement in defect detection capability.

  That is, in the defect inspection apparatus described in Patent Document 1, a defect is detected from each of a plurality of (multivalued) image data captured by an area sensor by the following image processing ([Patent Document 1 [[ 0032] to [0035]).

  First, the multivalued image data is binarized, and the white area and the black area are labeled as detection targets. Next, a white area having an area (number of pixels) exceeding a specified value (a relatively large value corresponding to the area of the linear light source image; for example, 2500 pixels) from the white area to be detected is regarded as a linear light source image. exclude. Similarly, a black area having an area exceeding a specified value (a relatively large value commensurate with the area of the background area) from the black area to be detected is regarded as a background area (an image of an area having no defect in the molded sheet). exclude. Further, the white area and the black area having an area smaller than a specified value (a relatively small value close to 1 pixel; for example, 9 pixels) are regarded as noise and excluded from the white area and black area to be detected. Then, the remaining area that is not excluded from the white area and black area to be detected is detected as a defect.

  However, since the defect inspection apparatus described in Patent Literature 1 detects a defect based on inversion of brightness and darkness, a defect with low contrast on image data may not be detected.

  Here, various defects to be detected in the present invention will be described. The present invention is mainly intended to detect defects (appearance defects) accompanied by minute irregularities (particularly, irregularities having a height of about several μm) on the surface of the molded sheet. Examples of defects with minute irregularities include, for example, minute irregularities generated on the surface of a molded sheet due to bubbles or foreign matters; dents (dents generated by being pressed at points); bent marks (“knic”) Indentation marks (referred to as “streaks”) generated by the transport roll generated when the molded sheet is transported by the transport roll. These defects with minute irregularities are very difficult to detect with a defect inspection apparatus using a conventional line sensor. The present invention is primarily aimed at detecting these types of defects.

  In the present specification, for the sake of convenience, fine irregularities are locally concentrated (the diameter of the convex portion is about 1 mm or less (about several pixels or less when the resolution of the imaging device is 200 μm / pixel)), for example, Bubbles, foreign matter, dents, and the like are called point defects, and those in which minute irregularities are connected in a line and have a size exceeding 1 mm are called line defects. A typical line defect such as a streak has a width exceeding about 10 mm (a width exceeding about several tens of pixels when the resolution of the imaging device is 200 μm / pixel), typically about several tens of cm, In some cases, the width exceeds several tens of centimeters. The nick has a width of about 10 mm or less (about several tens of pixels or less when the resolution of the imaging apparatus is 200 μm / pixel), typically about a few mm, and has a point defect and a typical line defect. It has an intermediate property.

  Examples of defect images (moving images) captured by the imaging unit of the defect inspection apparatus described in Patent Document 1 are as shown in FIGS. In FIG. 14 to FIG. 15 of Patent Document 1, five consecutive frames constituting a moving image are shown as (a) to (e) in order of time. If this moving image is a normal moving image for television, the time interval (frame rate) between frames is 1/30 second. The time interval between frames depends on the characteristics of the imaging unit. In the defect inspection apparatus disclosed in Patent Literature 1, the imaging unit moves the imaging sheet from the imaging unit toward the center of the imaging region of the molded sheet (a rectangle indicated by a broken line on the surface of the molded sheet in FIG. 1 of Patent Literature 1) and the molded sheet. The illumination image (reflected image of the linear light source) is included in part of the imaging area, and there are areas without the illumination image (background area) on both sides of the illumination image in the imaging area. Are arranged to be. Therefore, the upward direction in the moving image shown in FIGS. 14 to 15 of Patent Document 1 coincides with the conveying direction of the molded sheet. In the imaging region of the molded sheet, the defect moves in the conveying direction of the molded sheet. Therefore, in the moving image shown in FIGS. 14 to 15 of Patent Document 1, the defect moves from bottom to top (illumination). The image is shown as a white banded area).

  FIG. 14 of Patent Document 1 is an example of a moving image of an imaging region (size in a direction orthogonal to the transport direction: 5 mm) including bubbles (point defects) imaged by the imaging unit. I can see it. In this movie, bubbles are not seen in the first frame (a), but in the second frame (b) where the point defect is close to the illumination image, bubbles begin to appear and the bubble is located at the edge of the illumination image 3 In the second frame (c), the bubbles can be seen relatively well, and in the fourth and fifth frames (d) and (e) where the bubbles have entered the illumination image, the bubbles are buried in the illumination light and cannot be seen.

  FIG. 15 of Patent Document 1 is an example of a moving image of an imaging region (size in a direction orthogonal to the conveyance direction: 5 mm) including a nick imaged by the imaging unit. However, as the knick passes, the illumination image that should have been originally reflected as a rectangular white area is distorted with time.

  FIG. 15 of Patent Document 1 is an example of a moving image of an imaging region (size in a direction orthogonal to the conveyance direction: 200 mm) including streaks (line defects) imaged by the imaging unit. In the first to third frames (a) to (c), the illumination image is somewhat distorted in an arcuate shape, but this degree of arcuate distortion is not caused by a defect and is pulling the molded sheet. This is what happens. On the other hand, in the fourth frame, the illumination image is relatively distorted in an S shape. The illumination image is distorted more than this level because there is a streak at the distorted position. Therefore, it is desired to detect a portion where the illumination image is distorted more than this level as a defect.

  However, the defect inspection apparatus of Patent Document 1 separates a bright area (a linear light source image and a defective area inside the linear light source image) and a dark area (a background area and a defective area inside the background area) in binarization processing. It has been found that a defect with low contrast on image data cannot be detected depending on the threshold value.

  That is, in the defect inspection apparatus of Patent Document 1, when the defect has a relatively high contrast (brightness change due to the defect) on the image data, it is shown in the vertical direction (molded sheet conveyance direction) luminance profile of FIG. In addition, the luminance change due to the defect is observed to cross the threshold value of the binarization process. More specifically, the maximum points on both sides of the minimum point are larger than the luminance value of the minimum point corresponding to the defect observed as the dark region in the linear light source image (the valley portion in FIG. 3). It becomes smaller than the luminance value. Further, the threshold value of the binarization process is larger than the luminance value of the local maximum point (the mountain portion in FIG. 3) corresponding to the defect observed as a bright region outside the linear light source image, and both sides of the local maximum point. It becomes larger than the luminance value of the local minimum point. Therefore, a defect observed as a dark region in the linear light source image and a defect observed as a bright region outside the linear light source image are detected. Therefore, the defect inspection apparatus of Patent Document 1 can reliably detect a defect having a relatively high contrast.

On the other hand, in the defect inspection apparatus of Patent Document 1, when the defect has a low contrast on the image data, the relationship between the vertical luminance profile of the defective portion and the threshold value of the binarization processing is the vertical luminance in FIG. The relationship shown in the profile, that is, the relationship in which the luminance change due to the defect does not cross the threshold value of the binarization process. As a guideline, the following formula (luminance change amount due to defects) <{(luminance level of linear light source image area)
-(Luminance level of background area)} / 2 (1)
In the case of satisfying, the relationship shown in FIG. 4 may be obtained. In the case of the relationship shown in FIG. 4, the defect is missed.

  However, even when the defect contrast on the image data is low and the expression (1) is satisfied, the luminance change due to the defect crosses the threshold of the binarization process as shown in the vertical luminance profile of FIG. For observed cases, defects can be detected. Therefore, the defect inspection apparatus of Patent Document 1 can detect a defect depending on the relationship between the luminance change due to the defect and the threshold value of the binarization process even if the equation (1) is satisfied.

  Furthermore, the defect inspection apparatus of Patent Document 1 uses a moving image. In the moving image, the defect image approaches the linear light source image for each frame, passes through the linear light source image, and the linear light source. A movement away from the image is observed. Therefore, even if the contrast of the defect on the image data is low and satisfies the formula (1), even if one of the plurality of images constituting the moving image, the luminance change due to the defect and the threshold of the binarization process If there is an image whose relationship is shown in FIG. 5, the defect can be detected.

  As described above, in the defect inspection apparatus disclosed in Patent Document 1, the probability of missing a defect increases as the luminance change amount due to the defect decreases with Equation (1) as a boundary. Therefore, the defect inspection apparatus of Patent Document 1 has room for improvement with respect to the certainty of detection of defects with low contrast.

  The present invention has been made in view of the above problems, and an object of the present invention is to provide a molded sheet defect inspection apparatus that can detect various defects more reliably.

  A defect inspection apparatus according to the present invention is a defect inspection apparatus that detects a defect of a molded sheet in order to solve the above-described problem, and a plurality of two-dimensional image data obtained by capturing a two-dimensional image of the molded sheet a plurality of times. Imaging means for generating a linear light source for illuminating the molded sheet so that an image of the linear light source is projected onto a part of the imaged region of the molded sheet, and the line on the molded sheet At least one of the molded sheet and the linear light source intersects the longitudinal direction of the linear light source and is orthogonal to the thickness direction of the molded sheet so that the position at which the image of the linear light source is projected changes. Moving means for moving in a direction, and line defect detection means for detecting a line defect from a plurality of two-dimensional image data generated by the imaging means, wherein the line defect detection means adds the two-dimensional image data to the two-dimensional image data. A line defect detection algorithm for fitting an edge of an image of a linear light source with a function curve and detecting a point where the distance between the edge of the image of the linear light source and the function curve is equal to or greater than a first threshold as a line defect, or About the edge of the image of the linear light source in the two-dimensional image data, the curvature in the vicinity region of each pixel is obtained, and the line defect is detected by the line defect detection algorithm that detects the portion where the curvature is equal to or greater than the second threshold as the line defect. It is characterized by being.

  According to said structure, a line defect can be detected more reliably.

  The defect inspection apparatus preferably further includes point defect detection means for detecting point defects from a plurality of two-dimensional image data generated by the imaging means. Thereby, not only a line defect but a point defect can be detected.

  In the defect inspection apparatus, the point defect detection means represents a change in luminance depending on a position along a straight line in the two-dimensional image data as a luminance profile, and a moving time between plots of the luminance profile plot group is constant. The brightness value of the target plot is calculated from the velocity vector of the mass point between the two plots immediately before the target plot and the acceleration vector of the mass point between the three plots immediately before the target plot. A point defect detection algorithm for predicting and detecting a point defect where a difference between the predicted luminance value and the actual luminance value is equal to or greater than a third threshold is detected as a point defect, or the two-dimensional image data is smoothed and smoothed The difference between the two-dimensional image data and the original two-dimensional image data is obtained as difference image data. A point defect is detected by a point defect detection algorithm that detects a point where the degree value is equal to or greater than the fourth threshold value and a point where the luminance value is equal to or less than the fifth threshold value (the fifth threshold value is smaller than the fourth threshold value) as a point defect. It is preferable to detect. Thereby, compared with the technique of patent document 1, it becomes possible to detect more reliably a point defect (for example, defect which shows a small luminance change like FIG. 4) with a low contrast. Therefore, both point defects and line defects can be reliably detected.

  A defect inspection apparatus according to the present invention is a defect inspection apparatus that detects a defect of a molded sheet in order to solve the above-described problem, and a plurality of two-dimensional image data obtained by capturing a two-dimensional image of the molded sheet a plurality of times. Imaging means for generating a linear light source for illuminating the molded sheet so that an image of the linear light source is projected onto a part of the imaged region of the molded sheet, and the line on the molded sheet At least one of the molded sheet and the linear light source intersects the longitudinal direction of the linear light source and is orthogonal to the thickness direction of the molded sheet so that the position at which the image of the linear light source is projected changes. Moving means for moving in the direction, and point defect detecting means for detecting point defects from a plurality of two-dimensional image data generated by the imaging means, and the point defect detecting means adds the two-dimensional image data to the two-dimensional image data. The change in luminance depending on the position along a straight line is represented as a luminance profile, and two plots immediately before the plot of interest are assumed, assuming a mass point that moves the plot group of luminance profiles so that the movement time between plots is constant. The luminance value of the target plot is predicted from the velocity vector of the mass point in between and the acceleration vector of the mass point between the three plots immediately before the target plot, and the difference between the predicted luminance value and the actual luminance value is A point defect detection algorithm for detecting a point defect equal to or greater than 3 as a point defect, or smoothing the two-dimensional image data, and calculating the difference between the smoothed two-dimensional image data and the original two-dimensional image data as difference image data Where the luminance value in the difference image data is greater than or equal to the fourth threshold and the luminance value is the fifth threshold (the fifth threshold is It is characterized in that in order to detect the point defect portion smaller than the threshold) is less than 4 by the defect detection algorithm that detects a point defect.

  Thereby, compared with the technique of patent document 1, it becomes possible to detect more reliably a point defect (for example, defect which shows a small luminance change like FIG. 4) with a low contrast.

  It is preferable that the defect inspection apparatus further includes a line defect detection unit that detects a line defect from a plurality of two-dimensional image data generated by the imaging unit. As a result, not only point defects but also line defects can be detected.

  As described above, the present invention has an effect of providing a molded sheet defect inspection apparatus that can detect various defects more reliably.

[Embodiment 1]
An embodiment of the present invention will be described below with reference to the drawings.

  The defect inspection apparatus according to the present embodiment detects a defect of a molded sheet. The defect inspection apparatus according to the present embodiment is suitable for inspection of a light-transmitting molded sheet, particularly a molded sheet made of a resin such as a thermoplastic resin. As the molded sheet made of resin, for example, the thermoplastic resin extruded from the extruder is passed through the gaps between the rolls to be treated to impart smoothness and gloss to the surface, and taken up while being cooled on the transport roll by the take-up roll. What was shape | molded by this is mentioned. Examples of the thermoplastic resin applicable to the present embodiment include methacrylic resin, methyl methacrylate-styrene copolymer, polyolefin such as polyethylene and polypropylene, polycarbonate, polyvinyl chloride, polystyrene, polyvinyl alcohol, and triacetyl cellulose resin. Etc. The molded sheet may be composed of only one of these thermoplastic resins, or may be a laminate (laminated sheet) obtained by laminating a plurality of types of these thermoplastic resins. Moreover, the defect inspection apparatus according to the present embodiment is suitable for inspection of optical films such as polarizing films and retardation films, in particular, long optical films that are wound and stored and transported in a web shape. Further, the molded sheet may have any thickness, and even if it has a relatively thin thickness generally called a film, it has a relatively thick thickness generally called a plate. It may be.

  Examples of defects in the molded sheet include bubbles (such as those that occur during molding), fish eyes, foreign matter, tire marks, dents, scratches, and other point defects; nicks, streaks (such as those that occur due to differences in thickness), and the like. Can be mentioned.

  The configuration of the defect inspection apparatus 1 according to the present embodiment will be described below with reference to FIGS. 1 and 2. FIG. 1 is a functional block diagram showing the configuration of the main part of the defect inspection apparatus 1. FIG. 2 is a schematic diagram showing an overview of the defect inspection apparatus 1. In FIG. 2, the light and darkness of the surface of the molded sheet is shown inverted so that a member overlapping the molded sheet can be easily identified. Therefore, the black area on the surface of the molded sheet in FIG. 2 is actually a bright area, and the white area on the surface of the molded sheet in FIG. 2 is actually a dark area.

The defect inspection apparatus 1 transports a rectangular shaped sheet 2 in a fixed direction by a conveying device (moving means) 3 while n shaped sheets 2 illuminated by the linear light source 4 (n is an integer of 2 or more). The imaging units (imaging means) 5 _{1 to} 5 _n capture images a plurality of times to generate a plurality of two-dimensional image data in each of the imaging units 5 _{1 to} 5 _n , and analyze based on the generated plurality of two-dimensional image data. The apparatus 6 detects a defect of the molded sheet 2.

The defect inspection apparatus 1 includes a conveying device (moving means) 3 that conveys the molded sheet 2 and an imaging region in the molded sheet 2 (region imaged by the imaging units 5 _{1 to} 5 _n ; a broken line on the surface of the molded sheet 2 in FIG. A linear light source 4 for illuminating the molding sheet 2 so that an image of the linear light source 4 is projected on a part of the rectangular shape indicated by (2), and a reflection image (from the linear light source 4) of the linear light source 4 The direct light is reflected by the molding sheet 2 and formed as a result of reaching the imaging units 5 _{1 to} 5 _n ) and the reflection image of the molding sheet 2 (scattered light from the linear light source 4 is molded). Imaging units 5 _{1 to} 5 _n that capture a plurality of times a two-dimensional image including an image of the molded sheet 2 formed as a result of being reflected by the sheet 2 and reaching the imaging units 5 _{1 to} 5 _n, and a plurality of two-dimensional images Based on the image data, the image processing algorithm ( Recessed and a analyzer 6 for detecting defects of a molded sheet 2 by the detection algorithm).

  The conveying device 3 conveys the molded sheet 2 in a direction perpendicular to its thickness direction, particularly in its longitudinal direction so that the position on the molded sheet 2 where the image of the linear light source 4 is projected changes. . The conveying device 3 includes, for example, a sending roller and a receiving roller that convey the molded sheet 2 in a certain direction, and measures the conveying speed with a rotary encoder or the like. The conveyance speed is set to about 2 m to 12 m / min, for example. The conveyance speed in the conveyance device 3 is set and controlled by an information processing device or the like (not shown).

The linear light source 4 has a longitudinal direction in a direction intersecting with the conveying direction of the molded sheet 2 (for example, a direction orthogonal to the conveying direction of the molded sheet 2), and the reflected image of the linear light source 4 is a molded sheet. 2 are arranged so that there are regions (background regions) where there is no linear light source image on both sides of the reflected image of the linear light source 4 in the imaging region. The linear light source 4 is not particularly limited as long as it emits light that does not affect the composition and properties of the molded sheet 2. For example, a fluorescent light (especially a high frequency fluorescent light), a metal halide lamp, a halogen light is used. Such as a transmission light. Incidentally, the linear light source 4, sandwiching the molded sheet 2 is disposed at a position opposed to the image pickup unit 5 ₁ to 5 _n, direct light from the transmission image of the linear light source 4 (linear light source 4 is a molded sheet 2 An image of the linear light source 4 formed by passing through and reaching the imaging units 5 _{1 to} 5 _n ) and a transmission image of the molded sheet 2 (scattered light from the linear light source 4 passes through the molded sheet 2 and is captured) A two-dimensional image including an image of the molded sheet 2 formed by reaching the parts 5 _{1 to} 5 _n may be taken by the imaging parts 5 _{1 to} 5 _n .

Each of the imaging units 5 _{1 to} 5 _n captures a two-dimensional image including a reflected image of the linear light source 4 and a reflected image of the molded sheet 2 a plurality of times, and generates and outputs a plurality of two-dimensional image data. The imaging units 5 _{1 to} 5 _n include an area sensor including an imaging element such as a charge coupled device (CCD) or a complementary metal-oxide semiconductor (CMOS) that captures a two-dimensional image. The size of the defects detected by the defect inspection apparatus 1 is dependent on the resolution of the imaging unit 5 ₁ to 5 _n, may be selected the resolution of the imaging unit 5 ₁ to 5 _n to the size of the defect to be detected. Note that the three-dimensional shape (ratio of width to height) of the defect detected by the defect inspection apparatus 1 basically does not depend on the resolution of the imaging units 5 _{1 to} 5 _n , so that the camera resolution depends on the type of defect to be detected. It is not necessary to select

The imaging units 5 _{1 to} 5 _n are arranged such that the direction from the imaging units 5 _{1 to} 5 _n toward the center of the imaging region of the molded sheet 2 and the conveying direction of the molded sheet 2 form an acute angle. The imaging units 5 _{1 to} 5 _n have the entire area in the width direction of the molded sheet 2 (the direction orthogonal to the conveying direction of the molded sheet 2 and the direction orthogonal to the thickness direction of the molded sheet 2) of the imaging units 5 _{1 to} 5 _n . It arrange | positions along with the width direction of the shaping | molding sheet 2 so that it may image with at least one. By imaging the entire region in the width direction of the molded sheet 2 by the imaging units 5 _{1 to} 5 _n , defects in the entire region of the molded sheet 2 can be inspected.

The imaging interval (frame rate) of the imaging units 5 _{1 to} 5 _n may be fixed, may be changed by the user operating the imaging units 5 _{1 to} 5 _n itself, and the imaging unit 5 _1-5 an information processing device connected to _n (not shown; optional) it may be made changeable by the user operating the. In addition, the imaging interval of the imaging units 5 _{1 to} 5 _n may be a fraction of a second, which is a time interval for continuous shooting by a digital still camera, but in order to improve inspection efficiency, The interval is preferably, for example, 1/30 second, which is a frame rate of general moving image data.

Here, the distance (conveyance distance) by which the molding sheet 2 is conveyed during the period from when each of the imaging units 5 _{1 to} 5 _n captures _one two-dimensional image until the next two-dimensional image is captured is the molding It is set to at least 1 / m (m is 2 or more) of the length of the imaging region along the conveyance direction of the sheet 2. Thereby, a two-dimensional image including the same portion of the molded sheet 2 is captured m times. m is preferably sufficiently larger than 2. By increasing the number of times of imaging of the same location of the molded sheet 2, it is possible to inspect defects with high accuracy.

As shown in FIG. 1, the analysis device 6 receives a plurality of two-dimensional image data output from each of the imaging units 5 _{1 to} 5 _n , detects a defect based on the plurality of two-dimensional image data, and detects a detection result ( Line defect image analysis units (line defect detection means) 61 _{1 to} 61 _n and point defect image analysis units (point defect detection means) 62 _{1 to} 62 _n for outputting inspection results), and detection results (inspection results) A display unit 64 for displaying, and a control CPU 63 for controlling these units in an integrated manner.

A plurality of two-dimensional image data generated by each of the imaging units 5 _{1 to} 5 _n is input to the line defect image analysis units 61 _{1 to} 61 _n and the point defect image analysis units 62 _{1 to} 62 _n , respectively.

Each of the line defect image analysis units 61 _{1 to} 61 _n and the point defect image analysis units 62 _{1 to} 62 _n has a plurality of (plurality) in which the positions of the linear light source images on the molding sheet 2 differ depending on the line defect detection algorithm. A defect is detected from the two-dimensional image data (of the frame), and the result is output as an inspection result. Each of the point defect image analysis units 62 _{1 to} 62 _n detects defects from a plurality of (multiple frames) two-dimensional image data in which the positions of the linear light source images on the molded sheet 2 are different by a point defect detection algorithm. The result is output as an inspection result. The line defect image analysis units 61 _{1 to} 61 _n and the point defect image analysis units 62 _{1 to} 62 _n have defects based on a plurality of two-dimensional image data in which the positions of the linear light source images on the molding sheet 2 are different. Therefore, the defect can be detected more reliably than the conventional defect inspection apparatus.

The line defect detection algorithm and the point defect detection algorithm will be described later. The parameters of the line defect detection algorithm and the point defect detection algorithm may be fixed, but information processing connected to the line defect image analysis units 61 _{1 to} 61 _n and the point defect image analysis units 62 _{1 to} 62 _n. The device (not shown; can be omitted) may be changed by a user operation.

The line defect image analysis units 61 _{1 to} 61 _n are said to have a line defect when a line defect is detected from L (L ≦ m) or more of a plurality of two-dimensional image data by a line defect detection algorithm. The result may be output as an inspection result, and in other cases, a result indicating that there is no line defect may be output as an inspection result. From the L or more of a plurality of two-dimensional image data by a line defect detection algorithm When a line defect is detected, the information of the line defect position may be output as an inspection result, and in other cases, the inspection result may not be output. When the number m of 2D image data is 3 or more and L is 2 or more, if the number of 2D image data in which line defects are detected by the line defect detection algorithm is small, the line defect detection is performed. The result is regarded as false information (one that is not originally a defect is mistakenly detected as a defect) and is excluded. Thereby, false information can be reduced. In addition, when outputting the information of a line defect position as an inspection result, it is necessary to use the line defect detection algorithm which can obtain | require a line defect position.

The point defect image analysis units 62 _{1 to} 62 _n are said to have a point defect when a point defect is detected from L (L ≦ m) or more of a plurality of two-dimensional image data by a point defect detection algorithm. The result may be output as an inspection result, and in other cases, a result indicating that there is no point defect may be output as an inspection result. From a point defect detection algorithm, L or more of a plurality of two-dimensional image data may be output. When a point defect is detected, the point defect position information may be output as an inspection result, and in other cases, the inspection result may not be output. If the number m of 2D image data is 3 or more and L is 2 or more, if the number of 2D image data in which point defects are detected by the point defect detection algorithm is small, the point defect detection is performed. The result is regarded as false information (one that is not originally a defect is mistakenly detected as a defect) and is excluded. Thereby, false information can be reduced. In addition, when outputting the information of a point defect position as an inspection result, it is necessary to use the point defect detection algorithm which can obtain | require a point defect position.

The control CPU 63 integrates the inspection results output from the line defect image analysis units 61 _{1 to} 61 _n and the point defect image analysis units 62 _{1 to} 62 _n to obtain inspection result information corresponding to the entire region of the molded sheet 2. It is created and stored in a storage device (not shown) and displayed on the display unit 64. The inspection result information corresponding to the entire area of the molded sheet 2 includes information indicating whether or not the entire area of the molded sheet 2 has a defect, a defect map of the entire area of the molded sheet 2, and the like. When the inspection result information corresponding to the entire region of the molded sheet 2 is created, a defect is detected in at least one of the line defect image analysis units 61 _{1 to} 61 _n and the point defect image analysis units 62 _{1 to} 62 _n. If there is a defect, inspection result information is created assuming that a defect exists.

When the control CPU 63 creates a defect map of the entire area of the molded sheet 2 as inspection result information corresponding to the entire area of the molded sheet 2, the line defect image analysis units 61 _{1 to} 61 _n and the point defect image are used. Each of the analysis units 62 _{1 to} 62 _n converts the coordinate position on the two-dimensional image data into the coordinate position on the molding sheet 2 to generate defect position information, and outputs the defect position information to the control CPU 63. As coordinate conversion processing of each of the image analysis units 61 _{1 to} 61 _n for line defects and the image analysis units 62 _{1 to} 62 _n for point defects, for example, paragraphs [0037] to [0041] and [0050] of Patent Document 1 are used. The treatment described in [0053] can be used. The defect map information may be output to a marking device (not shown) and an information processing device (not shown), and the marking device may mark the defect position on the molded sheet 2 based on the defect map. This marking device has, for example, an arm provided along the width direction of the molded sheet 2 and a marker head having a pen or the like, and the marker head reciprocates on the arm in the width direction of the molded sheet 2. Thus, marking can be performed at an arbitrary position on the molded sheet 2. The information on the marked defect position can be used for, for example, a process of cutting the molded sheet 2 into a sheet of a predetermined size and then dividing the sheet into a normal product and a defective product.

In the above-described embodiment, the linear light source 4 is fixed and the molded sheet 2 is conveyed, but the position of the linear light source 4 on the molded sheet 2 where the image is projected changes. All you need is it. Therefore, the molding sheet 2 may be fixed and the linear light source 4 may be moved, or both the molding sheet 2 and the linear light source 4 may be moved in different directions or at different speeds. When the linear light source 4 is moved while the molding sheet 2 is fixed, it is preferable to move the imaging units 5 _{1 to} 5 _n in the same direction as the linear light source 4 at the same speed. Thereby, a plurality of two-dimensional image data including a linear light source image can be obtained. The method of moving the linear light source 4 while fixing the molded sheet 2 can avoid distorting the linear light source image by pulling the molded sheet 2 with the conveying device 3. Since the length is limited to the length corresponding to the movable range of the linear light source 4, in order to efficiently inspect the long molded sheet 2, the molded sheet 2 is used as in the above embodiment. It is preferable to carry.

In the above embodiment, the line defect image analysis units 61 _{1 to} 61 _n and the point defect image analysis units 62 _{1 to} 62 _n are two-dimensionally obtained from the same imaging units 5 _{1 to} 5 _n. Although the defect was detected based on the image data, the line defect image analysis units 61 _{1 to} 61 _n and the point defect image analysis units 62 _{1 to} 62 _n were obtained from different imaging units. A defect may be detected based on each. Thus, the imaging conditions of the imaging units 5 _{1 to} 5 _n (the distance from the molded sheet 2 and the angle between the conveying direction of the molded sheet 2 and the imaging direction) are set to conditions suitable for the defect to be detected. It becomes possible. The imaging unit for imaging the two-dimensional image data used by the line defect image analysis units 61 _{1 to} 61 _n is an imaging unit for imaging the two-dimensional image data used by the point defect image analysis units 62 _{1 to} 62 _n. It is preferable to increase the distance from the molded sheet and to narrow the angle between the conveying direction of the molded sheet 2 and the imaging direction. Thereby, since both a point defect and a line defect can be imaged on the optimal imaging conditions, both a point defect and a line defect can be detected still more accurately.

Further, in the above embodiment, the dispersion in the imaging section ₅ 1-5 image for 2-dimensional image of the line defect captured by the _n analysis unit ₆₁ 1 to 61 _n and the point defect image analysis unit ₆₂ 1 through 62 _n molded sheet from and had been treated, the imaging unit 5 ₁ to 5 _n captured based on the relative position between the two-dimensional image captured by the imaging unit 5 ₁ to 5 _n at the n pieces of 2-dimensional images Alternatively, one full-width image including the entire region in the width direction may be synthesized, and a defect may be detected by one line defect image analysis unit and one point defect image analysis unit based on the full-width image. As a method for synthesizing one full-width image from n two-dimensional images, for example, the method described in paragraph [0050] of Patent Document 1 can be used.

Next, a line defect detection algorithm and a point defect detection algorithm used in the line defect image analysis units 61 _{1 to} 61 _n and the point defect image analysis units 62 _{1 to} 62 _n will be described. As the line defect detection algorithm and point defect detection algorithm, the following seven types of defect detection algorithms A to G can be used.

[Defect detection algorithm A]
The defect detection algorithm A will be described below with reference to FIG. FIG. 6A is an example of multi-valued two-dimensional image data (hereinafter referred to as original image data) generated by one of the imaging units 5 _{1 to} 5 _n , and the upper side of the image is the downstream side in the transport direction. The lower side of the image is the upstream side in the transport direction. In FIG. 6A, a band-shaped white region extending in the center in the horizontal direction is a linear light source image, a dark region existing inside the linear light source image, and a small white region existing in the vicinity of the linear light source image. Is a defect.

In this defect detection algorithm A, the following processing is performed for each of a plurality of original image data generated by the imaging units 5 _{1 to} 5 _n .

  First, the original image data is converted into pixel column data (luminance value (pixel value) and position data; luminance profile; one-dimensional image data) for each row along the vertical direction (conveying direction of the molding sheet 2). To divide.

  Next, for the data of each pixel column, first edge determination processing for searching for an edge from one end (the upper end in FIG. 6A) to the other end (the lower end in FIG. 6A) as described below. I do. First, the second pixel from the one end side of the pixel column is set as the target pixel, and it is determined whether the luminance value of the target pixel is smaller than the threshold value T1 by the luminance value of the adjacent pixel adjacent to the target pixel on the one end side. If it is determined that the luminance value of the pixel of interest is smaller than the threshold value T1 by the luminance value of the adjacent pixel, it is determined that the adjacent pixel is the first edge, and the position of the first edge is recorded and processed. In the other cases, the luminance value of the target pixel is higher than the luminance value of the adjacent pixel in the determination while moving the target pixel one pixel toward the other end. The above determination is repeated until it is determined that the threshold value is smaller than the threshold value T1, and when it is determined that the luminance value of the target pixel is smaller than the threshold value T1 by the threshold value, the adjacent pixel is determined to be the first edge. Then, the position of the first edge is recorded, and the processing of the data of the pixel row to be processed is finished. Note that the threshold value T1 is an arbitrary natural number and may be the minimum unit of luminance values. When the threshold value T1 is the minimum unit of the luminance value, the above determination simply determines whether the luminance value of the target pixel is smaller than the threshold value than the luminance value of the adjacent pixel.

  Next, the following second edge determination process for searching for an edge from the other end to the one end is performed on the data of each pixel column as follows. First, the second pixel from the other end side is set as the target pixel, and it is determined whether the luminance value of the target pixel is larger than the threshold value T2 by the luminance value of the adjacent pixel adjacent to the other end side with respect to the target pixel. If it is determined that the luminance value of the pixel of interest is greater than the threshold value T2 by the luminance value of the adjacent pixel, the pixel of interest is determined to be the second edge, the position of the second edge is recorded, and the process In other cases, the luminance value of the target pixel is set to a threshold value that is higher than the luminance value of the adjacent pixel in the determination while moving the target pixel one pixel toward the other end. The above determination is repeated until it is determined that the pixel is larger than T2, and when it is determined that the luminance value of the pixel of interest is larger than the threshold value T2 by the luminance value of the adjacent pixel, it is determined that the pixel of interest is the second edge. Then, the position of the second edge is recorded, and the processing of the data of the pixel row to be processed is finished. The threshold value T2 is an arbitrary natural number and may be the minimum unit of luminance values. When the threshold value T2 is the minimum unit of the luminance value, the above determination simply determines whether the luminance value of the target pixel is greater than the threshold value by the luminance value of the adjacent pixel.

  An example of the first edge detected by the first edge determination process is indicated by “Δ” in FIG. 6A, and an example of the second edge detected by the second edge determination process is shown in FIG. This is indicated by “◯” in FIG. As can be seen from FIG. 6A, since there is no edge other than the edge of the linear light source image in the defect-free region, the first edge and the second edge are on the other end side in the linear light source image. The edges coincide with each other (the lower edge in the example of FIG. 6A) and coincide with each other. On the other hand, as can be seen from FIG. 6A, in the defective area (white area and black area), at least one of the first edge and the second edge coincides with the edge of the defective area, and the linear light source Since the edge is shifted to the edge search start side from the edge on the other end side in the image, the first edge and the second edge are separated from each other.

  Therefore, next, for the data of each pixel column, the distance (number of pixels) from the first edge to the second edge is obtained as the distance between the edges. FIG. 6B shows a profile in which the obtained distance between the edges is plotted with respect to the position of the pixel row (lateral coordinate). Then, if there is a pixel row whose edge distance is greater than or equal to the threshold T3, it is determined that there is a defect. The threshold T3 is an arbitrary natural number and may be one pixel. When the threshold value T3 is one pixel, a pixel column whose edge distance is not 0 is determined as a pixel column including a defect. The threshold value T3 may be determined as appropriate according to the size of an allowable defect. When the multi-value two-dimensional image data is two-dimensional image data having 256 gradations (brightness values 0 to 255; 8 bits), for example, 3 is preferred.

  In the example of FIG. 6, the upper end is set as one end (first edge search start side). However, it is arbitrary which end of the pixel column is set as one end, and the lower end may be set as one end. In that case, the first edge and the second edge coincide with the upper edge in the linear light source image in an area having no defect.

  This defect detection algorithm A can detect various point defects with a certain degree of certainty. However, the detection certainty of minute point defects such as bubbles and fish eyes is not so high. On the other hand, the defect detection algorithm A is not suitable for detecting line defects. This defect detection algorithm A is hereinafter referred to as “edge profile method 1”.

[Defect detection algorithm B]
The defect detection algorithm B fits the edge of the image of the linear light source in the two-dimensional image data with a function curve, and the distance between the edge of the image of the linear light source and the function curve is equal to or greater than a threshold value T5 (first threshold value). A part is detected as a defect.

The defect detection algorithm B will be described below based on FIG. FIG. 7A is an example of original image data generated by one of the imaging units 5 _{1 to} 5 _n , and the upper side of the image is the downstream side in the transport direction, and the lower side of the image is the upstream side in the transport direction. . In FIG. 7A, a band-like white region extending in the center lateral direction is a linear light source image, and a locally distorted portion (not smooth) of the lower edge of the linear light source image is a defect. .

In this defect detection algorithm B, the following processing is performed for each of a plurality of original image data generated by the imaging units 5 _{1 to} 5 _n .

  First, at least one of the edges of the linear light source image is obtained from the original image data. An example of the edge of the obtained linear light source image is indicated by “◯” in FIG. Although the lower edge of the linear light source image is obtained in the example of FIG. 7, the upper edge of the linear light source image may be obtained, or both the upper edge and the lower edge of the linear light source image may be obtained. .

  As a method for obtaining an edge of a linear light source image, a method of extracting an edge using a known edge extraction filter (for example, a Sobel filter) and using a strong edge as an edge of the linear light source image, two-dimensional image data Is divided into pixel row data for each row and then a strong edge is obtained from the data of each pixel row to obtain an edge of a linear light source image. The method described in [0057] of Patent Document 1 A method of obtaining a light source image region (a method of performing binarization and labeling, and extracting a region having an area larger than a predetermined value from the labeled regions as a linear light source image region), and the like. Here, as an example, a method will be described in which original image data is divided into pixel row data for each row, and then a strong edge is obtained from the data of each pixel row to obtain an edge of a linear light source image. First, the original image data is divided into pixel column data for each row along the vertical direction (conveying direction of the molded sheet 2). Next, processing for searching for the edge from the one end (the upper end in FIG. 7A) toward the other end (the lower end in FIG. 7A) is performed on the data of each pixel column. First, the second pixel from the one end side is set as the target pixel, and it is determined whether the luminance value of the target pixel is smaller than the threshold value T4 (T4 is a natural number) than the luminance value of the adjacent pixel adjacent to the one end side with respect to the target pixel. . In order to detect only strong edges, the threshold value T4 at this time is set to a relatively large value. If it is determined that the luminance value of the target pixel is smaller than the threshold value T4 by the threshold value of the adjacent pixel, it is determined that the adjacent pixel is an edge of the linear light source image, and the edge position of the linear light source image is recorded. Then, the processing of the data of the pixel row to be processed is finished. In other cases, the pixel of interest is compared with the luminance value of the adjacent pixel in the determination while moving the pixel of interest one pixel toward the other end. The determination is repeated until it is determined that the luminance value of the pixel of interest is smaller than the threshold value T4. If the luminance value of the pixel of interest is determined to be smaller than the threshold value T4 by the luminance value of the adjacent pixel, the adjacent pixel is a linear light source image. The edge position of the linear light source image is recorded, and the processing of the pixel row to be processed is completed.

  Next, the obtained sequence of edges of the linear light source image is fitted to a smooth curve represented by a function (fitting is performed using a function curve) to obtain a fitting curve (function curve). Examples of the function used for fitting include an n-order function (n is 2 or more), a Gaussian function, a Lorentz function, a Forked function, a combination of these functions, and the like. Is preferred. In addition, as a fitting evaluation method used for fitting, a least square method can be used, for example.

  Next, the two-dimensional image data is divided into pixel row data for each row along the vertical direction (conveying direction of the forming sheet 2), and for each pixel row data, from the fitting curve to the edge of the linear light source image. Is obtained as the fitting degree. FIG. 7B shows a profile in which the obtained fitting degree is plotted with respect to the position of the pixel row (coordinates in the horizontal direction (the direction perpendicular to the conveyance direction of the molded sheet 2 and the direction perpendicular to the thickness direction of the molded sheet 2)). ). If there is a pixel column having a fitting degree equal to or greater than the threshold value T5, it is determined that the edge position of the linear light source image in the pixel column is defective. Thereby, the presence or absence of a defect can be determined, and the position of the defect can also be obtained. As described above, it is possible to detect a line defect that appears as a local distortion of the edge of the linear light source image (a fine distortion of the linear light source image near the edge). The threshold T5 used for the determination is an arbitrary natural number and may be one pixel. In the case where the threshold value T5 is 1 pixel, if there is a pixel row whose fitting degree is not 0, it is determined that there is a defect. The threshold value T5 may be appropriately determined according to the size of an allowable defect, but is preferably 4 when the multi-valued two-dimensional image data is 256-gradation two-dimensional image data.

  In this defect detection algorithm B, in addition to determining the presence or absence of a defect, a defect position may be obtained. In that case, a pixel column having a fitting degree equal to or greater than the threshold value T5 may be extracted, and a pixel position between the edge of the linear light source image and the fitting curve in the extracted pixel column may be obtained as a defect position.

  This defect detection algorithm B can detect various line defects with high certainty. On the other hand, this defect detection algorithm B is not suitable for detecting point defects. This defect detection algorithm B is hereinafter referred to as “edge profile method 2”.

[Defect detection algorithm C]
The defect detection algorithm C smoothes the two-dimensional image data, obtains the difference between the smoothed two-dimensional image data and the original two-dimensional image data as difference image data, and the luminance value in the difference image data is the threshold value T6B (first 4 where the threshold value is 4 or higher; T6B is an arbitrary positive number) or higher, and the luminance value is lower than the threshold T6D (fifth threshold; T6D is an arbitrary positive number smaller than T6B) or lower. .

  The defect detection algorithm C will be described in further detail below. In this defect detection algorithm C, the contrast in the lateral direction of the linear light source image (the direction perpendicular to the conveying direction of the molded sheet 2 and the direction orthogonal to the thickness direction of the molded sheet 2) is compared with the brightness in the lateral direction due to the defect. Utilizing the fact that the change has a higher spatial frequency, a high frequency component of the original image data is extracted, and a portion whose luminance value in the high frequency component is equal to or higher than the threshold value T6B or lower than the threshold value T6D is detected as a defect.

  (1) First, the original image data is smoothed (smoothed) in the horizontal direction using a horizontal smoothing filter (matrix) of 1 row and n columns (n is an integer of 3 or more). As a result, the high frequency region of the horizontal luminance change in the original image data is removed, and only the low frequency component remains for the horizontal luminance change (the low frequency component of the horizontal luminance change and the vertical luminance change remain). As the horizontal smoothing filter, a weighted averaging filter such as a Gaussian filter, an averaging filter, or the like can be used. Note that n is preferably 3.

  (2) Next, the smoothed image data is subtracted from the original image data (the luminance value of each pixel is subtracted). Thereby, only the high frequency component of the horizontal luminance change in the original image data remains.

  (3) Next, the image data obtained by the subtraction is smoothed using a 3 × 3 pixel smoothing filter (operator). This smoothing removes noise and leaves high-frequency components other than noise. As the smoothing filter, it is preferable to use a smoothing filter that preserves edges, such as a bilateral filter or a median filter.

  (4) Next, from the original image data, an upper edge (an edge on the downstream side in the transport direction) and a lower edge (an edge on the upstream side in the transport direction) in the linear light source image are obtained. Since the method for obtaining the edge of the linear light source image is the same as that described with respect to the defect detection algorithm B, description thereof is omitted. Next, with the conveyance direction of the forming sheet 2 in the original image data as the X axis, the minimum value Min is obtained from the X coordinate values of all the pixels constituting the upper edge, and the X coordinate values of all the pixels constituting the lower edge are determined. The maximum value Max is obtained from the inside. Then, the value obtained by subtracting the maximum value Max from the minimum value Min is regarded as the width W of the linear light source image, and an area in which the X coordinate value is expanded outward by the width W from the area from the minimum value Min to the maximum value Max is inspected. Define as That is, an area where the X coordinate value is equal to or greater than Min− (Min−Max) and equal to or less than Max + (Min−Max) is defined as an inspection area. This process is for narrowing down the inspection target area to only the linear light source image and its vicinity area. In addition, the reason why the inspection area is expanded outside the area where the X coordinate value is from the minimum value Min to the maximum value Max is that the inspection area includes a slight distortion of the linear light source image.

  (5) Next, from the luminance value of the pixel in the inspection area in the image data (high-frequency component other than noise) after the smoothing of (3) above, the bright side (luminance Threshold T6B on the higher side) and threshold T6D on the dark side (lower luminance side).

T6B = (average luminance value in inspection area) + (standard deviation of luminance value in inspection area) × k
T6D = (average luminance value in the inspection area) − (standard deviation of luminance values in the inspection area) × k
(K represents a positive parameter)
Note that the value of k may be determined as appropriate according to the allowable defect size, and is, for example, 1.5, 3, 4.5, or the like.

  (6) Next, a process for determining whether the luminance value of all pixels in the inspection area in the image data subjected to the smoothing of (3) is equal to or higher than the threshold value T6B or lower than the threshold value T6D (threshold value). Process), and pixels having a threshold value T6B or more or a threshold value T6D or less are extracted as defective portions. Thereby, the presence or absence of a defect can be determined, and the position of the defect can also be obtained.

Incidentally, if the noise contained in the original image data generated by the imaging unit 5 ₁ to 5 _n is small, it may be omitted smoothing process (3). If it is not necessary to limit the inspection target area to only the linear light source image and its neighboring area, the process of defining the inspection area in (4) is omitted, and the processes in (5) and (6) are performed on the entire image data. You may go to

  The defect detection algorithm C can detect all point defects including minute defects such as bubbles and fish eyes with high certainty. On the other hand, the defect detection algorithm C is not suitable for detecting line defects. However, regarding the processing time, the defect detection algorithm other than this defect detection algorithm C is shorter (the processing time of the defect detection algorithm C is, for example, about 40 ms per frame). This defect detection algorithm C is hereinafter referred to as “high-pass filter method”.

[Defect detection algorithm D]
The defect detection algorithm D will be described below with reference to FIGS. FIG. 8 is an example of original image data generated by one of the imaging units 5 _{1 to} 5 _n , and the upper side of the image is the downstream side in the transport direction, and the lower side of the image is the upstream side in the transport direction. In FIG. 8, a strip-like white region extending in the horizontal direction in the center is a linear light source image, and a dark region existing inside the linear light source image and a small white region existing in the vicinity of the linear light source image are defective. It is. In FIG. 8, the curves above and below the linear light source image indicate the upper and lower limits of the inspection target region.

In the defect detection algorithm D, the following processing is performed for each of the plurality of original image data generated by the imaging units 5 _{1 to} 5 _n .

  First, the original image data is divided into pixel row data for each row along the vertical direction (conveying direction of the molding sheet 2), and a plot row representing a change in luminance value depending on the position in each pixel row is vertical. Obtained as a directional luminance profile. An example of the obtained vertical luminance profile is shown in FIG. This example is a vertical luminance profile related to the pixel row at the position indicated by the arrow in FIG. 8, and y indicates the downward direction (the direction indicated by the arrow in FIG. 8; the direction opposite to the conveying direction of the molded sheet 2). Y coordinate.

  Next, the depth of the valley portion (see FIG. 8) is obtained for the vertical luminance profile of each pixel column. That is, first, regarding the vertical luminance profile of each pixel row, all local maximum points and local minimum points are obtained, and for all the obtained local minimum points, the luminance value (local minimum value) of the local minimum point and the local minimum point are the most. The difference from the brightness value (maximum value) of the nearest local maximum point is obtained as the depth of the valley portion. If the depth of the obtained valley portion is equal to or greater than a threshold value T7 (T7 is a positive number), it is determined that there is a defect. The threshold value T7 may be determined as appropriate according to the allowable defect size, but is preferably 0.25 × 255, for example, when the multivalued two-dimensional image data is 256-gradation two-dimensional image data. .

  This defect detection algorithm D has a relatively short processing time. This defect detection algorithm D can detect various point defects with a certain degree of certainty. In particular, it is suitable for detecting a point defect that causes local light / dark reversal near the edge of a linear light source image. However, the image including the point defect and the vicinity thereof needs to have a high contrast, and the certainty of detecting a small point defect such as a bubble, fish eye, or a tire mark is not so high. On the other hand, this defect detection algorithm D is not suitable for detecting line defects. This defect detection algorithm D is hereinafter referred to as “peak method”.

[Defect detection algorithm E]
The defect detection algorithm E represents a change in luminance depending on a position along a straight line in the two-dimensional image data as a luminance profile, and a plot point of the luminance profile is a mass point that moves so that the movement time between plots is constant. Assuming that the luminance value of the attention plot is predicted from the velocity vector of the mass point between the two plots immediately before the attention plot and the acceleration vector of the mass point between the three plots immediately before the attention plot, and the predicted luminance value And a difference between the actual luminance value and a threshold value T8 (third threshold value; T8 is a natural number) or more is detected as a defect.

  The defect detection algorithm E will be described below with reference to FIGS. This defect detection algorithm E improves the accuracy of the peak method, and detects a defect based on the difference between the actually measured value and the predicted value instead of the valley depth.

In this defect detection algorithm E, the following processing is performed for each of a plurality of original image data generated by the imaging units 5 _{1 to} 5 _n .

  First, similarly to the peak method, the vertical luminance profile of each pixel column is obtained. An example of the obtained vertical luminance profile is shown in FIG. 10 with the luminance value as the x-axis. A circled portion in the vertical luminance profile is a profile corresponding to a defect to be detected by the defect detection algorithm E.

  Next, regarding the vertical luminance profile of each pixel column, a mass point moving from one end of the plot column to the other end is assumed so that the movement time between adjacent plots is constant regardless of the distance between plots. . Then, it is assumed that the mass point moves from the plot c to the adjacent plot b, from the plot b to the adjacent plot a, and from the plot a to the adjacent plot d as shown in FIG. Further, it is assumed that the plot d is a plot corresponding to the target pixel.

  Then, the velocity vector and acceleration vector of the mass point in the three plots a to c where the mass point passed immediately before the plot d are obtained. That is, the velocity vector of the mass point in the section from plot b to plot a based on the travel time and the coordinates of the two plots a and b (x coordinate, y coordinate) that the mass point passed immediately before plot d. Ask for. Further, the velocity vector of the mass point in the section from the plot c to the plot b is obtained based on the moving time and the coordinates of the plots b and c (x coordinate, y coordinate) through which the mass point passed immediately before the plot d. Based on the velocity vector of the mass point in the interval from plot b to plot a and the velocity vector of the mass point in the interval from plot c to plot b, the acceleration vector of the mass point in the interval from plot c to plot a Ask for. Then, the coordinates (position) of the plot d are predicted from the velocity vector of the mass point in the section from the plot b to the plot a and the acceleration vector of the mass point in the section from the plot c to the plot a.

  The difference between the x-coordinate (luminance value) of the plot d predicted in this way and the actual (actually measured) x-coordinate (luminance value) of the plot d is obtained, and if these differences are greater than or equal to the threshold T8, the plot d The pixel corresponding to is extracted as a defective part. Thereby, the presence or absence of a defect can be determined, and the position of the defect can also be obtained. The threshold value T8 may be determined as appropriate according to the allowable defect size, but is preferably 20 when the multi-valued two-dimensional image data is 256-gradation two-dimensional image data.

  This defect detection algorithm E can detect various point defects with high certainty. This defect detection algorithm E is hereinafter referred to as “peak method 2”.

[Defect detection algorithm F]
The defect detection algorithm F will be described below with reference to FIG. FIG. 12A is an example of original image data generated by one of the imaging units 5 _{1 to} 5 _n , and the upper side of the image is the downstream side in the transport direction, and the lower side of the image is the upstream side in the transport direction. . In FIG. 12A, a band-like white region extending in the horizontal direction at the center is a linear light source image, and a locally distorted portion (a place where the inclination with respect to the horizontal line is large) of the lower edge of the linear light source image. It is a defect.

In the defect detection algorithm F, the following processing is performed for each of the plurality of original image data generated by the imaging units 5 _{1 to} 5 _n .

  First, at least one of the edges of the linear light source image is obtained from the original image data. An example of the edge of the obtained linear light source image is indicated by “◯” in FIG. Although the lower edge of the linear light source image is obtained in the example of FIG. 12, the upper edge of the linear light source image may be obtained, or both the upper edge and the lower edge of the linear light source image may be obtained. . Since the method for obtaining the edge of the linear light source image is the same as that described with respect to the defect detection algorithm B, description thereof is omitted.

  Next, with the horizontal direction as the x-axis and the vertical direction as the y-axis, the curve of the edge of the linear light source image (edge profile) y = f (x) is second-order differentiated to obtain a second-order differential profile. An example of the obtained secondary differential profile is shown in FIG.

  Then, for each pixel at the edge of the linear light source image, it is determined whether the secondary differentiation is greater than or equal to a threshold T9 (T9 is a positive number), and a pixel (at a high frequency) whose secondary differentiation is greater than or equal to the threshold T9 is defective. Determined as a site. Thereby, the presence or absence of a defect can be determined, and the position of the defect can also be obtained. The threshold value T9 may be appropriately determined according to the allowable defect size.

  This defect detection algorithm F is suitable for detecting a line defect that appears as a local curvature of an edge of a linear light source image. This defect detection algorithm F does not have a high defect detection capability. This defect detection algorithm F is hereinafter referred to as “edge curve method 1”.

[Defect detection algorithm G]
The defect detection algorithm G obtains the curvature in the vicinity region of each pixel (range of 2n + 1 pixels in the vicinity) for the edge of the image of the linear light source in the two-dimensional image data, and the curvature is a threshold T10 (second threshold; T10 is positive) (Number) or more are detected as defects.

  The defect detection algorithm G will be described below with reference to FIG.

In this defect detection algorithm G, the following processing is performed for each of a plurality of original image data generated by the imaging units 5 _{1 to} 5 _n .

  First, at least one of the edges of the linear light source image is obtained from the original image data. An example of the edge of the obtained linear light source image is shown in FIG. Since the method for obtaining the edge of the linear light source image is the same as that described with respect to the defect detection algorithm B, description thereof is omitted.

  Next, the curvature at each point (each pixel) is determined for the edge curve of the linear light source image. The method of obtaining the curvature is not particularly limited, and may be a method of calculating using mathematically determined mathematical formulas. However, in such a method, the processing time becomes long. It is preferable to obtain the curvature approximately.

  (1) A range consisting of n pixels (white dots in FIG. 13) left and right (or front and back) with respect to the target pixel (black point in FIG. 13) on the edge and a target pixel (range of 2n + 1 pixels in the vicinity of the target pixel) Is a calculation target range (n is a natural number). n may be appropriately determined according to the size of the allowable defect, but is preferably 30, for example. The example of FIG. 13 is a case where n is 3.

  (2) Next, the pixels at both ends of the calculation target range are connected by a straight line.

  (3) The predicted luminance value is obtained from the straight line over all pixels in the calculation target range, the increment of the actual luminance value (luminance value on the edge curve) with respect to the predicted luminance value is obtained, and the increment or the absolute value of the increment is calculated. Accumulate. With the integrated value obtained here, the curvature in the range of 2n + 1 pixels in the vicinity of the target pixel can be sufficiently approximated (a curvature value substantially equivalent to the curvature calculated using mathematical formulas can be obtained). Here, in the configuration using the incremental integrated value, when a slight change in luminance value occurs on the straight line or below the straight line within the calculation target range as shown in FIG. 13C. Therefore, the approximate value of the curvature is obtained by canceling and ignoring these changes. On the other hand, in the configuration using the integrated value of the absolute value of the increment, even when such a change occurs, an approximate value of the curvature including such a change is obtained. As shown in FIG. 13C, if a minute change in luminance value that goes up or down the straight line within the calculation target range is detected as a defect, a configuration using an integrated value of the absolute value of the increment is used. And it is sufficient. On the other hand, if such a change is allowed and it is not desired to detect it as a defect, an incremental integrated value may be used.

  (4) The integrated value is calculated for all the pixels on the edge while moving the pixel of interest one pixel at a time from the end of the edge in the linear light source image. Thereby, a profile (curvature profile) of an approximate value of curvature is generated.

  Next, for each pixel at the edge of the linear light source image in the curvature profile, it is determined whether the curvature is equal to or greater than a threshold value T10, and a pixel having a calculated curvature equal to or greater than the threshold value T10 is defined as a defective portion (or defect candidate). judge. Thereby, the presence or absence of a defect can be determined, and the position of the defect can also be obtained. Since the edge of the linear light source image is slightly bent because the molded sheet 2 is somewhat warped, it should be allowed as a non-defect if the curvature of the edge of the linear light source image is up to a certain extent. Therefore, the threshold value T10 should be relatively large. The threshold value T10 may be determined as appropriate according to the allowable defect size, but is preferably 110, for example, when the multi-valued two-dimensional image data is 256-gradation two-dimensional image data.

  This defect detection algorithm G can detect various line defects with high certainty. This defect detection algorithm G is hereinafter referred to as “edge curve method 2”.

In the present embodiment, the combination of the line defect detection algorithm and the point defect detection algorithm respectively used in the line defect image analysis units 61 _{1 to} 61 _n and the point defect image analysis units 62 _{1 to} 62 _n is any of the following: .

(A) The line defect detection algorithm used in the image analysis units 61 _{1 to} 61 _n for line defects is the edge profile method 2 or the edge curve method 2, and the point defect detection algorithm used in the image analysis units 62 _{1 to} 62 _n for point defects. Is the high-pass filter method or peak method 2.

(B) The line defect detection algorithm used in the line defect image analysis units 61 _{1 to} 61 _n is the edge profile method 2 or the edge curve method 2, and the point defect detection algorithm used in the point defect image analysis units 62 _{1 to} 62 _n. Is a defect detection algorithm other than the high-pass filter method or the peak method 2.

(C) The line defect detection algorithm used in the line defect image analysis units 61 _{1 to} 61 _n is a defect detection algorithm other than the edge profile method 2 or the edge curve method 2, and the point defect image analysis units 62 _{1 to} 62 _n The point defect detection algorithm to be used is the high-pass filter method or the peak method 2.

  Of the combinations (A) to (C), the combination (A) is most preferable. In the case of the combination (A), both line defects and point defects can be reliably detected. In the case of the combination (B), the line defect can be detected reliably. In the case of the combination (C), the point defect can be reliably detected.

[Embodiment 2]
The following will describe another embodiment of the present invention with reference to FIG. For convenience of explanation, members having the same functions as those shown in the first embodiment are denoted by the same reference numerals, and description thereof is omitted.

The defect inspection apparatus according to the present embodiment has the same configuration as the defect inspection apparatus 1 according to the first embodiment except that the analysis apparatus 6A illustrated in FIG. 14 is provided instead of the analysis apparatus 6 illustrated in FIG. ing. As shown in FIG. 14, the analysis device 6 _</ b _{> A} is obtained by omitting the point defect image analysis units 62 _{1 to} 62 _n from the analysis device 6 shown in FIG. 1.

In this embodiment, the line defect detection algorithm used in the line defect image analysis units 61 _{1 to} 61 _n is the edge profile method 2 or the edge curve method 2. In this embodiment, a line defect can be reliably detected.

  In addition, although the defect inspection apparatus of this embodiment may be used independently, it is preferable to use it combining with the defect inspection apparatus which can detect a point defect. As a result, not only line defects but also point defects can be inspected. The defect inspection apparatus capable of detecting point defects combined with the defect inspection apparatus of the present embodiment may be various known defect inspection apparatuses, but may be a defect inspection apparatus of the third embodiment described later. preferable. Thereby, both a line defect and a point defect can be inspected reliably.

[Embodiment 3]
The following will describe another embodiment of the present invention with reference to FIG. For convenience of explanation, members having the same functions as those shown in the first embodiment are denoted by the same reference numerals, and the description thereof is omitted.

The defect inspection apparatus according to the present embodiment has the same configuration as the defect inspection apparatus 1 according to the first embodiment except that the analysis apparatus 6B illustrated in FIG. 15 is provided instead of the analysis apparatus 6 illustrated in FIG. ing. As shown in FIG. 15, the analysis device 6 </ b _{> B} is obtained by omitting the line defect image analysis units 61 _{1 to} 61 _n from the analysis device 6 shown in FIG. 1.

In this embodiment, the point defect detection algorithm used in the point defect image analysis units 62 _{1 to} 62 _n is the high-pass filter method or the peak method 2. In this embodiment, a point defect can be reliably detected.

  In addition, although the defect inspection apparatus of this embodiment may be used independently, it is preferable to use it combining with the defect inspection apparatus which can detect a line defect. As a result, not only point defects but also line defects can be inspected. The defect inspection apparatus capable of detecting line defects combined with the defect inspection apparatus of the present embodiment may be various known defect inspection apparatuses, but is preferably the defect inspection apparatus of the second embodiment. Thereby, both a line defect and a point defect can be inspected reliably.

[Experimental example]
Next, in order to confirm the effect of the present invention, results of an experiment conducted using 14 types of experimental defect inspection apparatuses similar to the defect inspection apparatus according to the above embodiment will be shown.

In the first to seventh experimental defect inspection apparatuses, the imaging units 5 _{2 to} 5 _n are omitted from the defect inspection apparatus according to the third embodiment, and instead of the conveyance roll, the molding sheet 2 is used as the conveyance device 3 on the surface thereof. The conveyor which carries on board and uses is used. The first to seventh experimental defect inspection apparatuses are for detecting point defects from a sample including point defects.

The first experimental defect inspection apparatus includes point defect image analysis units 62 _{1 to} 62 _n using the edge profile method 1, and the second experimental defect inspection apparatus uses a point defect image using the edge profile method 2. comprising an analysis unit 62 ₁ through 62 _n, the third experimental defect inspection apparatus includes a defect image analysis unit 62 ₁ through 62 _n that using a high-pass filter method, the fourth experimental defect inspection apparatus peak method includes a defect image analysis unit 62 ₁ through 62 _n points using, fifth experimental defect inspection apparatus includes a defect image analysis unit 62 ₁ through 62 _n that using peak method 2, sixth experiment The defect inspection apparatus for use includes point defect image analysis units 62 _{1 to} 62 _n using the edge curve method 1, and the seventh experimental defect inspection apparatus uses the edge curve method 2 (method using an incremental integrated value). gastric point defect image analysis unit ₆₂ 1 through 62 _n Eteiru.

The eighth to fourteenth experimental defect inspection apparatuses omit the imaging units 5 _{2 to} 5 _n from the defect inspection apparatus according to the second embodiment, and serve as the conveyance device 3 in place of the conveyance roll and the molded sheet 2 on the surface thereof. The conveyor which carries on board and uses is used. The eighth to fourteenth experimental defect inspection apparatuses are for detecting a line defect from a sample including the line defect.

The eighth experimental defect inspection apparatus includes line defect image analysis units 61 _{1 to} 61 _n using the edge profile method 1, and the ninth experimental defect inspection apparatus uses a line defect image using the edge profile method 2. Analyzing units 61 _{1 to} 61 _n are provided, the tenth experimental defect inspection apparatus is provided with line defect image analyzing units 61 _{1 to} 61 _n using a high-pass filter method, and the eleventh experimental defect inspection apparatus is a peak method. Line defect image analysis units 61 _{1 to} 61 _n using the first defect, and the twelfth experimental defect inspection apparatus includes line defect image analysis units 61 _{1 to} 61 _n using the peak method 2, and the thirteenth experiment. The defect inspection apparatus for a line includes image analysis units 61 _{1 to} 61 _n for line defects using the edge curve method 1, and the fourteenth experimental defect inspection apparatus uses an edge curve method 2 (a method using an incremental integrated value). There was a line defect image analysis unit 61 ₁ It has a 61 _n.

  In this experimental example, as the molded sheet 2, ten types of polarizing film samples containing different types of point defects and six types of polarizing film samples containing different types of line defects were used. The ten types of samples containing point defects are: a sample 01 containing bubbles, a sample 02 containing fish eyes, a sample 03 containing a first foreign material, a sample 04 containing a second foreign material different from the first foreign material, a first sample Sample 06 including tire marks, Sample 07 including second tire marks different from the first tire marks, Sample 08 including first dents, Sample 09 including second dents different from the first dents Sample 11 including the first wound and Sample 12 including the second wound different from the first wound. Six types of samples including line defects include a sample 10 including a nick (line defect), a sample 13 including a second nick (line defect) different from the first nick, and streaks along the conveyance direction of the molded sheet 2. A sample 51 including a strong streak orthogonal to the conveying direction of the molded sheet 2, a sample 53 including a weak streak orthogonal to the conveying direction of the molded sheet 2, and an oblique direction with respect to the conveying direction of the molded sheet 2 This is a sample 53 containing streaks.

Further, in the first 1-14 laboratory defect inspection apparatus, as the image pickup unit 5 _1, by capturing a two-dimensional image, it is possible to generate a two-dimensional image data of horizontal 512 pixels × vertical 480 pixels 256 gradations A progressive scan area sensor using a CCD element was used. In the first to fourteenth defect inspection apparatuses, a high-frequency fluorescent lamp with a sharp edge hood (a hood that sharpens the edge of the linear light source image) was used as the linear light source 4. In the first to fourteenth defect inspection apparatuses for experiment, the conveying speed of the molded sheet 2 by the conveyor was 20 mm / sec (= 1.2 m / min). The position in the (front end of FIG. 2) at a distance of 145mm edge of the conveyor is set to be the center of the imaging region by the imaging unit 5 _1. In addition, a metal scale was used to measure 145 mm, and the distance from the metal scale to the defect was set to 55 mm.

In the first to seventh experimental defect inspection apparatuses for detecting point defects, the imaging region (field of view) on the molded sheet 2 is lateral (perpendicular to the conveying direction of the molded sheet 2 and in the thickness direction of the molded sheet 2). orthogonal directions) 51.2 mm × vertical (so that the conveying direction) 48 mm of the forming sheet 2, adjusting the position and angle of the imaging unit _{5 1.} However, among the 512 pixels × 480 pixels of the progressive scan area sensor, 512 pixels arranged in the lateral direction at the position of the 240th pixel from the top (vertical center position) are 51.2 mm wide on the surface of the molding sheet 2. adjusting the position and angle of the imaging unit 5 ₁ so as to photograph the area. That is, the distance from the forming sheet 2 to the imaging unit 5 ₁ adjusts the position of the imaging unit 5 ₁ so that 190 mm, also from the center of the condenser lens of the imaging unit 5 _first imaging direction (the imaging unit 51, angle between direction) toward the center of the area to be imaged by the imaging unit 5 ₁ a molded sheet 2 surface was adjusted angle of the imaging unit 5 ₁ so that 40 degrees. In this case, the imaging unit _{5 1} resolution is 100 [mu] m / pixel. In the first to seventh experimental defect inspection apparatuses for detecting point defects, as a progressive scan area sensor, the focal length is 25 mm, the minimum F value is 1.4, and the lens tip work distance is 270 mm. A C-mount lens mounted on a progressive scan area sensor body was used, and the aperture was adjusted to about 11.

  Further, in the first to seventh experimental defect inspection apparatuses for detecting point defects, the longitudinal direction of the linear light source 4 is orthogonal to the conveying direction of the molded sheet 2 and the distance from the molded sheet 2 to the linear light source 4 The linear light source 4 is arranged so that the straight line connecting the center of the imaging region on the molded sheet 2 and the center of the linear light source 4 forms an angle of 37 degrees with respect to the surface of the molded sheet 2. did.

In a 8-14 laboratory defect inspection apparatus for detecting a line defect, the imaging area on the molded sheet 2 so that the horizontal 204.8Mm × vertical 192 mm, to adjust the position and angle of the imaging unit 5 ₁ . However, among the 512 pixels × 480 pixels of the progressive scan area sensor, 512 pixels arranged in the lateral direction at the position of the 240th pixel from the top (vertical center position) are 204.8 mm wide on the surface of the molding sheet 2. adjusting the position and angle of the imaging unit 5 ₁ so as to photograph the area. That is, the distance from the forming sheet 2 to the imaging unit 5 ₁ adjusts the position of the imaging unit 5 ₁ so that 400 mm, also an angle of 15 degrees between the imaging direction of the imaging unit 5 ₁ a molded sheet 2 surface adjusting the angle of the imaging unit 5 ₁ so that. In this case, the imaging unit _{5 1} resolution is 200 [mu] m / pixel. In the eighth to fourteenth defect inspection apparatuses for detecting line defects, as a progressive scan area sensor, the focal length is 25 mm, the minimum F value is 1.4, and the lens tip work distance is 490 mm. Using a C-mount lens mounted on a progressive scan area sensor body, the aperture was adjusted to about 5.6-8.

  In the eighth to fourteenth experimental defect inspection apparatuses for detecting a line defect, the linear light source is arranged such that the longitudinal direction of the linear light source 4 forms an angle of 25 degrees with respect to the conveying direction of the molded sheet 2. 4 and the work distance of the linear light source 4 was set to 900 mm.

  Here, parameters of the defect detection algorithm are set so that a point defect having a diameter of 0.5 mm can be reliably detected. In the first and eighth experimental defect inspection apparatuses, the threshold value T3 in the edge profile method 1 is set to 3. In the second and ninth experimental defect inspection apparatuses, the threshold value T5 in the edge profile method 2 is set to 4. In the third and tenth experimental defect inspection apparatuses, k in the edge profile method 2 is set to 4.5, and the horizontal direction smoothing filter used for the edge profile method 2 is a 1 × 3 smoothing filter. In the fourth and eleventh experimental defect inspection apparatuses, the threshold value T7 in the peak method is set to 25% (255 × 0.25) of the maximum luminance value. In the fifth and twelfth experimental defect inspection apparatuses, the threshold value in the peak method 2 was set to 20. In the sixth and thirteenth experimental defect inspection apparatuses, the distance, which is a parameter used in the edge curve method 1, is set to 15, and k is set to 5. In the seventh and fourteenth experimental defect inspection apparatuses, the method of obtaining the curvature approximately as described above as the edge curve method 2 was used, and the calculation target range was set to a range of 30 pixels before and after the target pixel.

  And it is investigated whether a point defect can be detected from 10 types of samples containing a point defect using the 1st-7 experimental defect inspection apparatus, and a line defect is used using the 8-14 experimental defect inspection apparatus. It was investigated whether line defects could be detected from 6 types of samples including. The obtained results are shown in Table 1.

In the first to seventh experimental defect inspection apparatus, the total number of frames of the moving image formed sheet 2 to be imaged (all capture number) is a 150 sheets, one defect detected points or point defects imaging unit 5 ₁ It enters the imaging area within a period of up to exiting from the imaging area of the imaging unit 5 ₁ a, 30 frames of moving image is captured by the imaging unit 5 _1. That is, if a defect is always visible, among the 2-dimensional image data captured by the imaging unit 5 _1, the defect in a two-dimensional image data of 30 frames captured in the period (30 frames) image Will be included. In each row of point defects in Table 1, the number of frames (frames) in which defects are detected by each experimental defect inspection apparatus (defects are detected) out of 30 frames of two-dimensional image data captured within the above period. In the column “Number of detected frames”.

In the 8-14 laboratory defect inspection apparatus, within a period of up to one defect detected points leaves the imaging region of the imaging unit 5 ₁ enters the imaging area of the imaging unit 5 _1, 300 frames of video There is imaged by the imaging unit 5 _1. In each row of line defects in Table 1, the number of frames (frames) in which defects are detected by each experimental defect inspection apparatus among the 300 frames of two-dimensional image data captured within the above period is expressed as “number of detected frames”. ”Column.

  In Table 1, when the number of detected frames is 0, it indicates that a defect could not be detected, and the size of detected frames indicates the accuracy or certainty of defect detection. For example, when the number of detected frames is as small as 1 to 2 frames, it is considered that the accuracy of defect detection is low. In Table 1, the accuracy of defect detection is determined based on the determination criteria uniquely determined by the applicant of the present application, “×” (the number of detected frames is 0 frame), “Δ” (the number of detected frames is 1 to 2 frames), “ ”(Number of detected frames is 3 to 6 frames) and“ ◎ ”(number of detected frames is 7 frames or more), and are described in the“ judgment ”column.

  Further, in Table 1, the row of “false information” indicates how many frames of false information have occurred in 1800 frames (frames).

From the results in Table 1, it can be seen that all types of point defects can be detected with high accuracy by the defect inspection apparatus (third and fifth experimental defect inspection apparatuses) using the high-pass filter method and the peak method 2. That is, in the defect inspection apparatus according to the above embodiment, if the point defect detection algorithm used in the point defect image analysis units 62 _{1 to} 62 _n is the high-pass filter method or the peak method 2, all types of point defects are accurately detected. It can be seen that it can be detected well.

Further, from the results shown in Table 1, all types of line defects can be detected with high accuracy by the defect inspection apparatus (second and seventh experimental defect inspection apparatuses) using the edge profile method 2 and the edge curve method 2. I understand. That is, in the defect inspection apparatus according to the above-described embodiment, if the line defect detection algorithm used in the line defect image analysis units 61 _{1 to} 61 _n is the edge profile method 2 or the edge curve method 2, all types of line defects are obtained. It can be seen that can be accurately detected.

  In most cases shown in Table 1, the number of detected frames is smaller than the maximum value (30 for point defects and 300 for line defects). It can be seen that there is a possibility that the defect cannot be detected by the method of detecting the defect based only on one piece of the two-dimensional image data (moving image data). For example, one piece of two-dimensional image data is randomly selected from a plurality of pieces of two-dimensional image data used for defect detection in this experimental example, and the sample 13 is based on only one piece of the two-dimensional image data. When an attempt is made to detect the included nicks, the nicks can be detected only with a low probability of 8/300 even if a defect detection algorithm (edge curve method 2) that can detect the nicks most reliably is used.

  The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope shown in the claims, and embodiments obtained by appropriately combining technical means disclosed in different embodiments. Is also included in the technical scope of the present invention.

It is a functional block diagram which shows the structure of the principal part of the defect inspection apparatus which concerns on one Embodiment of this invention. It is a schematic diagram which shows the general view of the said defect inspection apparatus. It is a figure for demonstrating the problem of a prior art. It is a figure for demonstrating the problem of a prior art. It is a figure for demonstrating the problem of a prior art. It is a figure for demonstrating an example (edge profile method 1) of a defect detection algorithm. It is a figure for demonstrating the other example (edge profile method 2) of a defect detection algorithm. It is a figure for demonstrating the further another example (peak method) of a defect detection algorithm. It is a figure for demonstrating the further another example (peak method) of a defect detection algorithm. It is a figure for demonstrating the further another example (peak method 2) of a defect detection algorithm. It is a figure for demonstrating the further another example (peak method 2) of a defect detection algorithm. It is a figure for demonstrating the further another example (edge curve method 1) of a defect detection algorithm. It is a figure for demonstrating the further another example (edge curve method 2) of a defect detection algorithm. It is a functional block diagram which shows the structure of the principal part of the defect inspection apparatus which concerns on other embodiment of this invention. It is a functional block diagram which shows the structure of the principal part of the defect inspection apparatus which concerns on further another embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Defect inspection apparatus 2 Molded sheet 3 Conveyance apparatus (moving means)
4 linear light sources 5 _{1 to} 5 _n imaging unit (imaging means)
6 Analysis Device 6A Analysis Device 6B Analysis Device 61 _{1 to} 61 Image analysis unit for _n- line defects (line defect detection means)
62 _{1 to} 62 _n point defect image analysis unit (point defect detection means)

Claims (3)

A defect inspection device for detecting defects in a molded sheet,
Imaging means for imaging a two-dimensional image of the molded sheet a plurality of times to generate a plurality of two-dimensional image data;
A linear light source for illuminating the molded sheet so that an image of the linear light source is projected onto a part of the imaged region of the molded sheet;
At least one of the molded sheet and the linear light source intersects with the longitudinal direction of the linear light source so that the position of the image of the linear light source on the molded sheet changes, and the molded sheet Moving means for moving in a direction perpendicular to the thickness direction of
Line defect detection means for detecting a line defect from a plurality of two-dimensional image data generated by the imaging means,
The line defect detecting means is
With respect to the edge of the image of the linear light source in the two-dimensional image data, the curvature in the vicinity region of each pixel is obtained, and a line defect is detected by a line defect detection algorithm that detects a point where the curvature is equal to or greater than the second threshold as a line defect. Is what
Point defect detection means for detecting point defects from a plurality of two-dimensional image data generated by the imaging means ;
The point defect detecting means is
The change in luminance depending on the position along a straight line in the above two-dimensional image data is expressed as a luminance profile, and the plot of interest is assumed assuming a mass point that moves the plot group of luminance profiles so that the movement time between plots is constant. The luminance value of the attention plot is predicted from the velocity vector of the mass point between the two plots immediately before and the acceleration vector of the mass point between the three plots immediately before the attention plot, and the predicted luminance value and the actual luminance value are estimated. A point defect detection algorithm for detecting a point defect as a point defect that is equal to or greater than a third threshold, or
The two-dimensional image data is smoothed, a difference between the smoothed two-dimensional image data and the original two-dimensional image data is obtained as difference image data, and a luminance value in the difference image data is a fourth threshold value or more and A defect inspection characterized in that a point defect is detected by a point defect detection algorithm for detecting a point having a luminance value equal to or lower than a fifth threshold (the fifth threshold is smaller than the fourth threshold) as a point defect. apparatus. A defect inspection device for detecting defects in a molded sheet,
Imaging means for imaging a two-dimensional image of the molded sheet a plurality of times to generate a plurality of two-dimensional image data;
A linear light source for illuminating the molded sheet so that an image of the linear light source is projected onto a part of the imaged region of the molded sheet;
At least one of the molded sheet and the linear light source intersects with the longitudinal direction of the linear light source so that the position of the image of the linear light source on the molded sheet changes, and the molded sheet Moving means for moving in a direction perpendicular to the thickness direction of
Point defect detection means for detecting point defects from a plurality of two-dimensional image data generated by the imaging means,
The point defect detecting means is
The change in luminance depending on the position along a straight line in the above two-dimensional image data is expressed as a luminance profile, and the plot of interest is assumed assuming a mass point that moves the plot group of luminance profiles so that the movement time between plots is constant. The luminance value of the attention plot is predicted from the velocity vector of the mass point between the two plots immediately before and the acceleration vector of the mass point between the three plots immediately before the attention plot, and the predicted luminance value and the actual luminance value are estimated. A defect inspection apparatus, wherein a point defect is detected by a point defect detection algorithm for detecting, as a point defect, a portion having a difference between and a third threshold value or more. The defect inspection apparatus according to claim 2 , further comprising a line defect detection unit that detects a line defect from a plurality of two-dimensional image data generated by the imaging unit.

Images