JP4739044B2 - Appearance inspection device - Google Patents

Description

The present invention relates to a visual inspection apparatus for performing an appearance inspection of the outer peripheral surface of the cylindrical body.

  In the technical field of nuclear fuel, in order to maintain the quality of cylindrical bodies such as end plugs, rods, pipes, fuel rods, pellets, etc., appearance inspection of the cylinders is performed. Cylindrical bodies that easily reflect light such as metal surfaces are also used in general industries. Defects that are generally required to be inspected to maintain the quality of these cylinders are (1) cracks, chips, (2) dents, deformations, pits, voids, (3) burrs, scratches, wrinkles, scratches. , Hit mark, die mark, (4) unground, abnormally finished surface, (5) discoloration, uneven color, rough surface, rough skin, earthworm, oil burn, (6) foreign matter entrainment, fixed matter, (7) lint, dust , (8) deposits, fingerprints, dirt, stains, oil, (9) coating abnormality, coating cracking, (10) gloss unevenness, corrosion, rust, etc.

  Conventionally, the appearance of cylindrical bodies such as end plugs, rods, pipes, and fuel rods has been visually inspected and visually inspected using an automatic inspection device. In the visual appearance inspection, visual inspection by an inspector is performed by handling by hand or rolling on a surface plate. Also, in the inspection using the automatic inspection device, by rotating the cylindrical body, illuminating the imaging inspection part uniformly with epi-illumination, etc., imaging the rotating cylindrical body with the line sensor, and detecting the abnormal part We are inspecting.

  Cylindrical surface inspection of drums, metal bars, etc. is known as a general cylindrical body inspection using an automatic inspection device. In the field of nuclear fuel, bar inspection, inspection of tubes such as cladding tubes In addition, pellet outer appearance inspection and the like are known.

  By the way, the conventional visual appearance inspection is not stable because many inspections are performed by a person, and often depends on the experience of the operator. In addition, there is a problem that the burden on the worker is increasing due to aging and the like, and a problem that the cost is high.

  In addition, visual inspection in conventional automatic inspection equipment can be automated for specific defects, but the S / N (Signal to Noise Ratio) deteriorates for various types of defects, and is a level that can replace visual visual inspection. In order to achieve this, there is a problem that a large number of cameras must be arranged under various conditions, and a large-scale system must be made by using many image devices. In the inspection at a level replacing the visual appearance inspection, it is difficult to realize an automatic device that can be covered by economical cost and space.

  So far, the following techniques have been proposed for appearance inspection of a cylindrical body by an automatic inspection apparatus. First, Non-Patent Document 1 proposes a technique for detecting a scratch on the surface of a bar by a laser rotating operation method. In this technique, the appearance is inspected with a change in the amount of light at a certain position of speckle of laser light irradiated from a certain angle in the axial direction of a cylindrical bar-shaped article.

  Patent Document 1 proposes a technique for detecting a regular reflection light of a rotating pellet with a linear image sensor and extracting a defect by passing a video signal through a plurality of threshold value binarization circuits. In this technique, the illumination and the linear image sensor have an imaging arrangement in which the illumination and the linear image sensor are arranged in the same direction and the vertical direction, and the illumination and the linear image sensor are inspected in different positions.

  Patent Document 2 proposes a cylindrical body appearance inspection device that rotates and moves a cylindrical body, reads a one-dimensional image signal in the center of a subject of a captured image, and obtains a developed image. In this technique, a plurality of pellets are sequentially fed and inspected continuously.

  Patent Document 3 proposes a technique for an imaging device for scratch inspection of a cylindrical surface. In this technique, a rubber roller is rotated, a line sensor camera (one-dimensional image pickup means) picks up an image from a right angle direction of illumination, and an appearance inspection is performed with a two-dimensional developed image at an intermediate brightness level.

  Patent Document 4 proposes a technique of an appearance inspection apparatus in which illumination that is perpendicular to a rotating cylindrical surface is arranged, and a plurality of line sensor cameras are arranged on a staggered pattern with the illumination interposed therebetween. In this technology, specularly reflected light can be received by a plurality of line sensor cameras on a staggered pattern, and inspection is performed in a state where a long cylindrical object is covered with high resolution.

  Patent Document 5 proposes a technique of an appearance inspection device for a cylindrical workpiece. In this technique, a fixed cylinder is imaged at a fixed interval by a two-dimensional imaging means with a two-dimensional imaging means, and image information of the entire circumference is obtained at intervals (6 divisions) where the inspection areas overlap to confirm the overlapping portion. A determination is made as to whether the appearance is good or bad.

  Moreover, in patent document 6, the technique regarding a rod-shaped article illumination device is proposed. In this technique, inspection is performed by uniformly illuminating the surface of a cylindrical bar-shaped article.

In the techniques shown in Patent Documents 1, 3, and 4, line-shaped illumination is used, and photographing is performed by a line sensor. On the other hand, in the techniques shown in Patent Documents 5 and 6, the appearance inspection is performed by dividing the circumference into several parts by a two-dimensional imaging unit using a technique for uniformly illuminating the surface.
JP-A-56-160645 JP-A-62-039746 JP-A-9-014942 JP-A-9-304287 JP 2002-350358 A Japanese Patent Laying-Open No. 2005-069695 “Automation of the latest visual inspection”, Mitsubishi Electric Technical Review 56, P69, August 28, 1986.

  However, the techniques disclosed in Non-Patent Document 1 and Patent Documents 1 to 6 can be used by adjusting so that a specific defect can be detected. There is a problem that it is not a realistic appearance inspection that can ensure the S / N of the detection.

  In the visual appearance inspection of the curved surface of a cylindrical body, since the object is held or rolled and inspected, the positional relationship between the object, eyes, and lighting is somewhat flexible, and the curved surface of the cylindrical surface can be viewed from various directions and angles. Inspected by illumination and line of sight. That is, as in the automatic inspection image with the line sensor, the positional relationship between the inspection position of the object, the line sensor optical system, and the illumination is not a single condition, but only the regular reflection of the illumination or an image at a certain angle is seen. Instead, images of various reflection angles on the curved surface enter the eye, and the inspection is performed while the position of the eye varies within a certain range.

  On the other hand, the positional relationship between the object, the camera corresponding to the eyes, and the illumination in the automatic inspection apparatus using the conventional line sensor is fixed, and the positional relationship between the object, the line sensor, and the illumination is fixed. It was intended to capture information and identify defects. The positional relationship is a single condition in many technologies, and at most two independent line sensors are used, and can respond to flexible changes such as visual appearance inspection. There is no problem.

  In addition, the apparatus of Patent Documents 5 and 6 that attempts to capture and inspect a curved surface as a real image as much as possible with diffused light is to be inspected by an apparatus that divides and images a peripheral surface with a two-dimensional image under uniform illumination. The normal surface texture is easily emphasized only by inspection with a real image of diffused light illumination, and at various natural defect levels, it is very difficult to automate the separation logic of the defective part and the normal surface. In the inspection other than the defect having a high S / N in the image, there is a problem that it cannot cope with the defect identification ability like the inspector like the visual appearance inspection.

  Therefore, in the conventional cylindrical body appearance inspection apparatus, although it functions for detection of specific defects, detection performance with various defects that can be replaced with visual inspection cannot be obtained, and stable practical automatic appearance inspection is performed. There is a problem that it is difficult.

The present invention has been made to solve the above-mentioned problems, and its purpose is to perform a visual inspection that enables a variety of inspections to be performed with a simple device under a number of conditions that are closer to visual inspection in the inspection of a cylindrical body. To provide an apparatus .

The present invention is an appearance inspection apparatus for inspecting a cylindrical surface of a cylindrical body inspection sample, the rotating unit for rotating the cylindrical body inspection sample, and a cylindrical surface of the cylindrical body inspection sample on the cylindrical body axis. An illumination unit arranged so as to show a parallel striped pattern, and the cylindrical body axis and the optical axis are arranged at right angles, and a region of the cylindrical surface including the striped pattern is set in a predetermined circumferential direction. A camera that captures images at a speed corresponding to the resolution and outputs image data, and extracts predetermined line pixel data from the image data of a plurality of images output from the camera, and reconstructs and expands the extracted line pixel data Based on the image processing unit that generates image data and the developed image data generated by the image processing unit, an abnormal portion of the cylindrical surface of the cylindrical body inspection sample is detected, and the cylindrical body inspection sample is detected by the detected abnormal portion. Of cylindrical surface Comprising an abnormality checking unit for acceptability determination of the test sample by evaluating the normal, wherein the image processing unit, the developed image data, generates a plurality for each of the line pixel data, the plurality of developed image data generated by On the other hand, the pixel position is corrected according to the line of the pixel data of the extraction source so as to correspond to the position of the cylindrical surface, and the abnormality inspection unit is the same among the plurality of developed image data corrected by the image processing unit An abnormal part is detected by an abnormal part extraction filter from the plurality of developed image data by matching the positions .

  The present invention is the above-described invention, wherein the illuminating unit is arranged at a center of the camera so that the striped pattern appears in a captured image of the cylindrical surface of the camera parallel to the cylindrical body axis of the cylindrical body inspection sample. A plurality of sample axis parallel illuminations are provided on the left and right sides of the axis.

  In addition, according to the present invention, the appearance inspection apparatus is arranged so that a rotating unit that rotates the cylindrical body inspection sample and a striped pattern that is parallel to the cylindrical body axis appear on the cylindrical surface of the cylindrical body inspection specimen. A camera that is arranged so that the illumination unit, the cylindrical body axis and the optical axis are perpendicular to each other, captures an area of a cylindrical surface including a striped pattern at a speed corresponding to a predetermined circumferential resolution, and outputs image data An image processing unit that extracts predetermined line pixel data from the image data of a plurality of images output from the camera, reconstructs the extracted line pixel data to generate developed image data, and an image processing unit Based on the developed image data, the abnormal portion of the cylindrical surface of the cylindrical body inspection sample is detected, and the abnormal inspection of the cylindrical surface of the cylindrical body inspection sample is evaluated by the detected abnormal portion to determine pass / fail of the inspection sample. And a configuration comprising It was. Thereby, development image data of a wide range of illumination angles and imaging angles of the cylindrical body inspection sample can be obtained only by rotating the cylindrical body inspection sample once. Since this developed image data is image data including multiple conditions from bright field to dark field, the abnormal part is detected by various optimum abnormal part extraction filters that match the inspection request from the developed image data of a plurality of conditions, By determining with the optimum abnormal portion determination filter, it is possible to perform an automatic appearance inspection of the cylindrical surface under conditions closer to the visual inspection. In addition, since these are performed electronically, automatic appearance inspection with good reproducibility is possible with little inspection variation such as visual inspection.

In addition, the image processing unit according to the appearance inspection apparatus generates a plurality of developed image data for each line pixel data, and extracts pixel data corresponding to the position of the cylindrical surface with respect to the generated plurality of developed image data. The abnormality inspection unit matches the same position of the plurality of developed image data corrected by the image processing unit and corrects the abnormal portion by using the abnormal part extraction filter from the plurality of developed image data. It was set as the structure which performs a detection. As a result, an abnormal part that cannot be detected without using the developed image data of different illumination angles and imaging angles is mutually detected from the developed image data generated by rotating the cylindrical object specimen once. It becomes possible to detect by complementing. In addition, this makes it possible to further improve the S / N of the abnormal portion by using a plurality of developed image data having different matching illumination conditions and imaging angles.

  Further, according to the present invention, the illumination unit according to the appearance inspection apparatus sandwiches the central axis of the camera so that the striped pattern appears in the captured image of the cylindrical surface of the camera parallel to the cylindrical body axis of the cylindrical body inspection sample. It was set as the structure provided with the some sample axis parallel illumination arrange | positioned on the left and the right. Thereby, it becomes possible to project a striped pattern on the cylindrical surface of the cylindrical body inspection sample, and when extracting abnormal portions with developed image data of various imaging angles and multiple illumination angles from bright field to dark field, This is advantageous in that various detection conditions can be set.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
1 to 4 are views showing a cylindrical body appearance inspection apparatus 10 according to the present embodiment. FIG. 1 is a front view and an overall configuration diagram of a cylindrical body appearance inspection apparatus 10. FIG. 2 is a side view of a portion related to an apparatus for photographing the cylindrical body inspection sample 1 of the cylindrical body appearance inspection apparatus 10. FIG. 3 is a side view of the axial conveyance unit 9 of the cylindrical body appearance inspection apparatus 10. FIG. 4 is a schematic block diagram showing the internal configuration of the image processing apparatus 8.

  The cylindrical body inspection sample 1 to be inspected in the present embodiment is disposed on a pair of two rollers 61 of the rotating unit 6 as shown in FIG. The rotating unit 6 includes the two pairs of rollers 61 and a roller driving motor 62. The roller 61 is rotated at a constant peripheral speed by the roller driving motor 62, so that the cylindrical specimen 1 is rotated. Rotate smoothly at a constant speed. As shown in FIG. 1 and FIG. 2, the rotating unit 6 may be one when the cylindrical body inspection sample 1 is short, but when the cylindrical body inspection sample 1 is long, the rotation body 6 may be a cylindrical body inspection sample. An appropriate number corresponding to the length of 1 is arranged at a position where the cylindrical specimen 1 can be rotated stably. Further, when the cylindrical inspection sample 1 is long, a plurality of cylindrical appearance inspection devices 10 are arranged in alignment with the axis of the cylindrical inspection sample 1 within a range in which the cylindrical inspection samples 1 do not overlap each other, so that the inspection time can be reduced. There are also ways to earn.

  As shown in FIG. 3, the axial conveyance unit 9 includes a drive motor 91, an axial drive roller 92, and a drive roller lift-up function 93. The driving roller lift-up function 93 raises the cylindrical specimen 1 to a height that does not contact the rotating section 6 or places it on the roller 61 of the rotating section 6 so that the cylindrical specimen 1 rotates the axial driving roller 92. When it is done, it is lowered to a position where it does not touch the cylindrical specimen 1. The axial direction driving roller 92 is driven by the drive motor 91 after the cylindrical body inspection sample 1 is lifted by the driving roller lift-up function 93, thereby moving the cylindrical body inspection sample 1 to the next rotational imaging position. Let By repeating this operation, it is possible to inspect the appearance of a long rod-like cylinder. In addition, this feed information is also taken into the inspection information and can be used for specifying the defect position.

  The short cylindrical specimen 1 is set at the rotational imaging position by a pick-up mechanism of the cylindrical specimen 1 (not shown) and picked up after imaging all around. An abnormal part is detected from the image data of this all-round picked-up image, and the pass / fail determination of the cylindrical body inspection sample 1 is performed by comprehensive determination of the abnormal part on the entire surface. For the long rod-shaped cylindrical specimen 1 after the entire imaging of the imaging position is finished, the driving roller lift of the axial conveyance unit 9 installed at a necessary interval in consideration of the deflection of the cylindrical specimen 1 After being moved up by the up function 93 and moved by the axial drive roller 92, the drive roller lift up function 93 is lowered to the roller 61 of the rotating unit 6 so that it is sent to the next imaging position in the axial direction and arranged. Is done. In order to increase the inspection time, a plurality of cylindrical body appearance inspection devices 10 are arranged in alignment with the axis of the cylindrical body inspection sample 1 for the long rod-shaped cylindrical body inspection sample 1, and each is connected to a network. A comprehensive determination may be performed by a device that manages the plurality of cylindrical body appearance inspection devices 10 connected to the network.

  Next, as shown in FIGS. 1 and 2, in the camera 5 of the two-dimensional image sensor, the axis of the cylindrical body sample 1 of the rotating unit 6 and the image scanning line of the two-dimensional image sensor 52 are parallel, and two It arrange | positions so that the axis | shaft of the cylindrical body to-be-tested sample 1 may become a perpendicular direction with respect to the surface of the two-dimensional image sensor 52. FIG. The illumination unit 7 includes a sample axis parallel illumination 71, a sample axis parallel oblique illumination 72, a light shielding plate 73, and an illumination support 75 so that a striped pattern can be formed on the circumferential surface of the cylindrical specimen 1 to be inspected. Arranged at intervals. Here, the stripe pattern means that the light of the light shielding portion and the sparse sample axis parallel illumination 71 is reflected by the normal surface of the cylindrical surface and is captured by the camera 5 so as to appear in the captured image of the cylindrical surface. Specifically, the illumination support 75 is used to form a striped pattern having a brightness that is continuous with the captured image captured by the camera 5 with respect to the normal portion of the cylindrical surface of the cylindrical specimen 1 to be inspected. A sample axis parallel illumination 71 such as is arranged. At this time, an optical black-dyed light-shielding plate 73 is used so that it becomes dark during illumination, and members other than the illumination unit 7 perform an optical black-dyed finish to suppress light reflection, and unnecessary external light and wraparound are performed. The light is prevented from being reflected and reflected on the cylindrical surface of the cylindrical specimen 1.

  The sample axis parallel illumination 71 of the illuminating unit 7 is of a type, size, arrangement position, and brightness so that a striped pattern that can ensure the defect extraction stability required for the circumferential surface captured image of the cylindrical body inspection sample 1 can be obtained. Is determined. The type, size, location, and brightness are optimal based on the concentricity, outer diameter variation, bending, and abnormal part type and how much defect detection accuracy is required for the cylindrical specimen 1 Is determined.

  Further, the sample axis parallel illumination 71 reduces the oversight of directivity, so that the center axis of the camera 5, that is, the imaging of the camera 5, so that the fringes due to the reflected light appear parallel to the cylindrical body axis of the cylindrical body inspection sample 1. They are arranged on the left and right with the central axis of the lens 51 in between. Accordingly, circumferential oblique illumination from the left and right can be performed, and a developed image obtained by the illumination is an image in which the imaging circumferential angle and the illumination circumferential angle are diagonal. That is, an image viewed from the left and an image viewed from the right are obtained, and the positional relationship can be easily matched, so that the S / N in abnormality detection can be improved.

  In addition, the brightness of the sample axis parallel illumination 71 corresponds to the situation such as the normal part reflectivity of the circumferential surface of the cylindrical object to be inspected 1, so that the sample of the cylindrical object to be inspected 1 within the assumed range in which abnormality detection is performed. Is taken in advance, and each is set so as to have a brightness necessary for extracting an abnormal part with a good S / N.

  The sample axis parallel oblique illumination 72 is a position where the illumination light has a certain angle with respect to the cylindrical object 1 and the regular reflection light does not enter the camera 5 and has a weak normal surface. Thin stripes due to diffusely reflected light are arranged so as to appear parallel to the cylindrical axis of the cylindrical specimen 1. The sample axis parallel oblique illumination 72 is a line-shaped object when it is relatively close to the rotational imaging range, and when it can be placed at a position relatively far from the rotational imaging range, a projector from a small surface is used. It can also be used.

  In addition, the sample axis parallel oblique illumination 72 is disposed on the upper left and lower right of the axis in order to reduce oversight of directional defects. The arrangement of the defects in this way is when the illumination is from one direction with respect to a certain line of the cylindrical specimen 1 and is diagonally independent. This is because it is easy to improve the effect of the one-side axial oblique illumination detection. In addition, the sample axis parallel oblique illumination 72 is effective in detecting the circumferential grinding state of the cylindrical specimen 1, the dent-like abnormal part spreading around the circumference, the crack in the circumferential direction, and the abnormal part of the cylindrical surface. Demonstrate. In addition, when it is desired to increase the S / N of the abnormal portion extraction from the defect discrimination, the sample axis parallel oblique illumination 72 can be illuminated from four diagonal directions. Further, when sufficient abnormal part detection can be performed only by the sample axis parallel illumination 71, the sample axis parallel oblique illumination 72 can be omitted.

  FIG. 4 is a schematic block diagram showing the internal configuration of the image processing apparatus 8. As shown in FIG. 4, in the image processing apparatus 8, the condition information storage unit 101 stores in advance various setting data necessary for the inspection during the automatic inspection. The transmission / reception unit 81 is connected to the camera 5 and transmits an imaging condition, a data transmission condition, an imaging start signal, and an imaging end signal to the camera 5, and conversely, image data of an image captured from the camera 5. Receive. The operation unit 82 receives an operation of the operator of the image processing apparatus 8 and selects various modes such as an automatic inspection mode, a manual scanning mode, a sample evaluation test mode, and an imaging condition correction mode, and each mode test instruction / setting, and an imaging start instruction. Then, an imaging end instruction, a result transmission instruction, and the like are input. In addition, the camera 5 is (1) imaged full-screen image transfer, (2) image data transfer of the inspection portion, and (3) pixel line data transfer of the image pickup element of the camera 5 at a location where the developed image data is reconstructed. It is possible to transfer data by switching the above functions, and one of the data transfer methods is selected according to the data transmission condition considering the margin of the data transfer time, that is, the time allowed for the data transfer. Since this time occupies a large part of the inspection processing time, the following description will be given assuming that (2) image data transfer of the image of the inspection part is used in principle when time is allowed.

  The image storage unit 100 stores image data of a captured image received by the transmission / reception unit 81 from the camera 5. The image reconstruction unit 83 stores pixel data (hereinafter referred to as “pixel data”) of pixels having the same illumination condition and imaging angle from data of one or more rounds of captured images stored in the image storage unit 100. (Also described as line pixel data) is reconstructed to generate a plurality of developed image data. This one developed image data is obtained by taking an image of a cylindrical surface under a certain illumination condition at a constant imaging angle, with the circumferential surface of the cylindrical specimen 1 as a developed plane.

  When the analysis is performed based on the correlation between the plurality of developed image data, the correction unit 84 uses the correction information stored in advance in the condition information storage unit 101 to develop each of the developed image data reconstructed by the image reconstruction unit 83. The pixel relative position is corrected so that the pixel position corresponds to the same position on the surface of the cylindrical specimen 1.

The abnormality detection unit 87 detects an abnormal part of the developed image data based on the abnormal part extraction condition information preset in the condition information storage unit 101.
The abnormality determination unit 88 evaluates the entire surface based on the data detected as an abnormality by the abnormality detection unit 87 and determines whether the cylindrical body inspected sample 1 is rejected, suspended, or passed.

The output unit 85 is connected to the image monitor 200, outputs the captured image data received from the camera 5 by the transmission / reception unit 81 to the image monitor 200, and stores image data stored in the image storage unit 100 in response to a display mode instruction. The developed image data reconstructed by the image reconstruction unit 83, the developed image data corrected by the correction unit 84, the developed image data in which the extracted abnormal portion is marked, and the like are output.
The communication unit 86 is connected to a network or the like, and performs communication with the axial conveyance unit 9 and the rotation unit 6 such as preparation completion and sorting instructions, and information input of the cylindrical body sample 1 to be inspected. The copy data, image data, developed image data, and the like are transmitted to an external device.

  Here, a captured image captured by the camera 5 will be described. FIG. 5 shows a sample axis parallel illumination 71 and a sample axis parallel oblique angle placed on the illumination support 75 of the illumination unit 7 by placing the cylindrical body inspection sample 1 on the roller 61 with the roller drive motor 62 of the rotating unit 6 stopped. This is an image captured by the camera 5 from the vertical axis of the cylindrical object to be inspected 1 through the elongated observation window 76 with the illumination 72 dimmed to an appropriate brightness.

  Illumination light from the sample axis parallel illumination 71 is reflected on the normal surface of the cylindrical surface of the cylindrical object to be inspected 1, and the reflected light is formed into a cylindrical shape by the imaging optical path indicated by the dense two-dot chain line A in FIG. It reaches the camera 5 through the light shielding plate 53. The reflected light that reaches the camera 5 is collected by the imaging lens 51 and imaged by the two-dimensional imaging element 52. In the image picked up by the two-dimensional image pickup device 52 shown in FIG. 5, seven brightly reflected portions (reference numerals 301 to 307) indicate that the light of each sample axis parallel illumination 71 is the circumferential surface of the cylindrical specimen 1. 2 is a normal surface of the cylindrical surface of the specimen 1 to be inspected. The bottom stripe line (reference numeral 307) is the light from the sample axis parallel illumination 71 at the lower right end of FIG. The image was reflected and reflected.

  On the other hand, two portions (symbols 310 and 311) that are thinly reflected in the dark portion between the bright portions are light from the left and right sample axis parallel oblique illumination 72 in FIG. The diffuse reflection light of the normal part of the cylindrical ground surface of 1 is copied.

(Operation of cylindrical body appearance inspection apparatus 10)
Next, the operation of the cylindrical body appearance inspection apparatus 10 will be described with reference to FIG.
(Step S1: Setting)
First, the cylindrical body inspection sample 1 is placed on the roller 61.

(Step S2: lighting on / start imaging)
The image processing apparatus 8 turns on the sample axis parallel illumination 71 and the sample axis parallel oblique illumination 72 through the communication unit 86. As a result, a striped pattern appears on the image of the normal part of the circumferential surface of the cylindrical specimen 1 imaged by the camera 5.

(Step S3: Rotation of cylindrical object sample)
On the other hand, the image processing apparatus 8 drives the roller drive motor 62 through the communication unit 86 to rotate the roller 61 at a constant speed, thereby rotating the cylindrical body inspection sample 1 at a constant peripheral speed.

(Step S4: Shooting with a camera, capturing image data of a captured image)
When the cylindrical specimen 1 is rotated at a constant peripheral speed and is in an inspectable state, image data capture permission information of the captured image is transmitted to the image processing device 8 via the network. In the image processing apparatus 8 that has received the image data capture permission information, the transmission / reception unit 81 of the image processing apparatus 8 receives and captures a plurality of image data of one or more rounds captured by the camera 5 into the image storage unit 100. (Step S4). The image processing device 8 can always display the received image data on the image monitor 82 through the output unit 85 at any time, or can display the developed image data of the designated portion of the cylindrical specimen 1.

7 and 8 are diagrams showing captured images displayed on one image monitor 82 received from the camera 5.
FIG. 7 is a captured image in the case of using a plurality of sample axis parallel illuminations 71 and sample axis parallel oblique illuminations 72 as in FIG. 5 described above, and detection of abnormal portions is performed using developed image data of many imaging angles. In this case, it is advantageous to use a plurality of illuminations in that various detection conditions can be set when detecting an abnormality.
FIG. 8 is a captured image in the case where only one object with a large illumination tube is used for the sample axis parallel illumination 71, and sufficient abnormal part extraction necessary for inspection can be performed with developed image data with a small imaging angle. In addition, it is advantageous when it is necessary to stabilize the cylindrical body sample 1 when the outer diameter or the variation in bending is large. Further, in the cylindrical inspection sample 1 in which the imaging surface is not stabilized by the method of rotating on the roller 61 as shown in the figure, the imaging surface is stabilized above the portion that does not obstruct the imaging of the cylindrical inspection sample 1. It is advantageous to use a rotating part 6 of a type in which a guide roller that is not used is placed and the cylindrical specimen 1 is rotated along the guide roller.

  Here, the imaging time pitch required to obtain the circumferential resolution necessary for the inspection in step S4 will be described below. First, in normal appearance inspection, the same resolution is ensured in both the vertical and horizontal directions with respect to the cylindrical surface unless there is a special reason. Therefore, when the resolution of the image data captured by the camera 5 is 0.05 mm per pixel, the reconstructed developed image data for extracting an abnormal portion described later has a pixel pitch in the axial direction of 0.05 mm. The pixel pitch in the circumferential direction is also set to 0.05 mm. The two-dimensional image pickup device 52 of the camera 5 has a high-pixel, high-sensitivity, high-speed two-dimensional image pickup device and an optical system that can secure 0.05 mm per pixel as the image resolution in the axial direction in the image data to be taken. use. In addition, the imaging lens 51 is a lens having an appropriate angle of view that can secure the necessary brightness and depth of field with high resolution and low distortion. The camera 5 is installed so that the imaging scanning line of the two-dimensional imaging device 52 of the camera is aligned parallel and perpendicular to the axis of the cylindrical object 1 to be inspected.

  At this time, the time pitch for capturing the image data is set to a rotational speed at which the rotational circumferential pitch of the cylindrical specimen 1 is 0.05 mm. For example, when the sample outer shape of the cylindrical body inspection sample 1 is 10 mm, the circumference is 31.4 mm. Therefore, it is necessary to take an image of 628 lines in one round. When the imaging time pitch is 2 milliseconds, the rotational speed is 1.256 seconds / circumference and the peripheral speed is 25 mm / second. After stable rotation at this peripheral speed, it is only necessary to capture the image data of a necessary number of captured images of one or more laps by capturing images at a time pitch of 2 milliseconds.

  When detecting an abnormal part by matching the developed image data of multiple conditions, in order to cover the deviation of the image data of the picked-up image so that the continuous image of one round can be completely matched, it is about 1.4 rotations. Take an image. At this time, the captured image capturing time at one location is about 1.8 seconds. Note that if the cylindrical body inspection sample 1 and the cylindrical body appearance inspection apparatus 10 are stable and the images before and after one round coincide with each other, the image data for one round may be used.

(Step S5: Generation of developed image data)
Next, when image data of a plurality of captured images for one round or more is recorded in the image storage unit 100, the image reconstruction unit 83 stores a plurality of images necessary for the developed image in the image storage unit 100. A plurality of developed image data are generated by reconstructing line pixel data set in advance in the circumferential direction. In addition, the reconstructed image reconstruction can be sequentially performed at the time of capturing captured image data.

  Here, the presetting for generating the developed image data will be described. First, a matching adjustment sample having the same diameter as that of the cylindrical object to be inspected 1 with a mark such as a cross mark that makes the matching state clear is arranged on the roller 61, and image data is captured by the camera 5 while rotating at a constant speed. To do. After confirming the stability of the rotation with the image data for each rotation, the pixel data is reconstructed from each line of the image data captured in the image storage unit 100 to generate developed image data. The generated developed image data is displaced in the circumferential direction according to the position of the image of each line on the data. That is, the Y-coordinates that are the circumferential coordinates of the respective developed image data do not coincide with each other, and in order to match and match a plurality of developed image data, each developed image is a sample having the same diameter as the cylindrical object to be inspected 1. It is necessary to determine the relative shift amount of the image data and register it as the relative shift amount. The correction of the relative deviation of the cylindrical specimen 1 is corrected by the deviation set in advance by the correction unit 84, and a plurality of developed image data that can be matched in units of pixels can be generated.

  In the case of the captured image as shown in FIG. 7, the developed image data of all the pixel lines is generated. At this time, when the peripheral speed of the cylindrical body is adjusted so as to coincide with the imaging time pitch of the captured image and the one-line feed time of the captured image, and the image is captured while rotating at a constant peripheral speed, the developed image data is substantially captured. The image data is shifted by the amount corresponding to the line. It is possible to make the target points of all the displayed developed images coincide by correcting the line portion where the deviation has occurred by the correction unit 84. That is, these corrected developed image data are developed image data in which the target points match at the pixel level and the imaging conditions such as the illumination angle and the imaging angle are different.

  When higher-accuracy abnormal part detection is required, corrected unfolded image data is generated by complementing the surrounding pixels in units of sub-pixels with a necessary resolution. In principle, the accuracy of XY of the size corresponding to the pixel unit of the developed image data is the same, but if the resolution in the circumferential direction may be coarser, increase the rotation speed or slow down the imaging pitch time. Also good. In order to make the resolution in the circumferential direction finer, the rotation speed may be lowered or the imaging pitch time may be shortened. However, when the rotational speed, the imaging pitch time, or the like is changed, it is necessary to change the value of the deviation used for correction in accordance with the change.

  FIG. 9 is a diagram showing an example of a developed image displaying the developed image data of the same pixel line generated from the image data of the captured image of one or more rounds in FIG. In the order of the developed images 401 to 405, the developed image 401 is a developed image of a pixel line at the brightest place, the developed image 402 is a developed image of a pixel line at a slightly brighter place, and the developed image 403 is developed of a pixel line at an average brightness. An image and a developed image 404 are developed images of pixel lines in a slightly dark place, and a developed image 405 is a developed image of pixel lines in the darkest place. The average brightness is the average value of the brightness when the brightest and the darkest.

  This will be specifically described. FIG. 10 is an image obtained by enlarging a part of the captured image of FIG. 8, and is an image in which the reflected light from the sample axis parallel illumination 71 of the thicker illumination tube is reflected on the normal surface of the cylindrical body inspection sample 1. In the photographed image, the darkest pixel line successively exists in order from the brightest pixel line. The above-described developed images 401 to 405 are generated by extracting the pixel data of each pixel line from the time-series image, reconstructing the pixel data in the height direction in time series, and generating the developed image data. For example, an image generated by extracting the brightest pixel line in FIG. 10 is a developed image 401, and developed images 402 to 405 are generated for each pixel line in order of brightness. Further, in the case of a captured image as shown in FIG. 7, it is assumed that the imaging conditions are different, and developed image data of pixel lines designated in advance for bright fields, dark fields, intermediate brightness portions, etc. for each illumination. Thus, the abnormal portion can be detected and determined from the developed image data having different imaging angle of view and illumination conditions.

  Note that it is possible to generate and inspect developed image data of all line pixel data from image data of a captured image. However, at the level required in the normal inspection, the developed image data from the nearby line pixel data tends to give a similar result. In addition, because there is a need to shorten the calculation processing time for generating the unfolded image data from the relationship with the imaging time pitch and rotation time, the abnormal part is extracted by the following method according to the priority of the defect to be detected. The developed image data to be used may be limited by thinning out, and the limited developed image data, for example, about 5 to 20 thinned-out developed image data may be inspected.

  Here, a method for limiting the developed image data will be described. (Procedure 1) First, sample samples having various defects to be detected are placed on the roller 61. The transmission / reception unit 81 that captures an image with the camera 5 and receives the image data of the captured image as illustrated in FIG. 7 transmitted from the camera 5 records the image data in the image storage unit 100. (Procedure 2) Next, the image reconstruction unit 83 reconstructs pixels of each line of the image data of the captured image for one round stored in the image storage unit 100 to generate developed image data. At this stage, the developed image data is generated for all the pixel lines so that the abnormal portion extraction evaluation of the developed image can be performed with all the imaging angles and all the brightness conditions without using the thinned developed image data.

  Then, using the image data in which the defective part to be detected is clearly specified as a teacher signal, a plurality of developed image data are abnormally detected from the developed image data having a high S / N in the normal part of the developed image data and the abnormal part including the defect. Selected by the automatic analysis program as candidates for part detection, the parameters of the model program for abnormal part extraction determination are considered by the optimum condition analysis evaluation program, and processing time and redundancy are also considered. Find out. The condition information storage unit 101 stores the inspection condition parameters of the sample including the pixel lines in the image data of the developed image generation source at that time.

  For those that can secure a sufficient S / N for defect detection with the developed image data of one condition, the defect detection is performed using one developed image data under the condition.

  On the other hand, for a defect that cannot secure sufficient S / N with one image of developed image data, that is, a defect that may be erroneously recognized, or a defect that can recognize only a part of the defect, an abnormal part including the part of the defect Only reconstructs each developed image data of a plurality of pixel lines of the image data of the captured image, and corrects the misalignment of the developed image data. A combination of developed image data so that the detected defect matches the actual situation as much as possible, or a sufficient S / N, an optimum abnormal part extraction / determination filter based on a plurality of developed image data, and an optimum parameter thereof are determined, Analyzed. Then, a plurality of most suitable combinations, optimum abnormal part extraction / determination filters, and optimum parameters thereof are stored in the data data condition information storage unit 101.

For example, a defect that is recessed in the circumferential direction as detected from a pixel line at a position corresponding to the sample axis parallel oblique illumination 72 is a defect because only a part of the depression is generated in oblique illumination. Only a part of can be detected. On the other hand, it is often possible to detect a part of the other recess from the developed image data from the pixel line at a position corresponding to the diagonal sample axis parallel oblique illumination 72. When the developed image data from the pixel line at the position where dark field illumination is performed by the sample axis parallel illumination 71 is combined with this, the abnormal portion of the dent is often clearly recognized. By using the optimum abnormal part extraction / determination filter and its optimum parameters, it is possible to increase the probability that the surface abnormality can be judged as a defect due to a dent.
For example, a defect that is gently depressed is often indistinguishable from the normal surface by illumination from all directions. In many cases, illumination from one direction can detect only a part of the recess, and the defect is often underestimated and not rejected. On the other hand, in the developed image data at a plurality of dark field positions between the sample axis parallel illumination 71, a part of the abnormal portion of the data recess is often clearly recognized, and the abnormality detected from each of the developed image data. By comprehensively analyzing and evaluating the data of the portion, the surface abnormality can be recognized as a defect due to a dent, and the probability of clearly accepting the pass / fail decision can be increased.

  Further, for example, if the defect is something like dust, the developed image data of the dark field portion sandwiched between the bright field portion and the bright field portion is used. In the dark field, dust appears brighter than the normal part, and in the bright field, often appears darker than the normal part. The same applies to the dark field corresponding to the sample axis parallel oblique illumination 72. When the developed image data of these front and rear visual fields and the developed image data at a position corresponding to left and right oblique illumination are matched and extracted, it is possible to extract as a clear image of dust. If a plurality of developed image data having a large difference in imaging angle is used, it may be possible to determine these three-dimensional shapes.

  Using the optimum condition extraction computer program for selecting, analyzing and evaluating the developed image data as described above, the most suitable combination of single and plural developed image data, the optimum abnormal part extraction / determination filter, and the optimum parameter Inspection conditions are determined and stored in the condition information storage unit 101. The conditions to be adopted are determined by the balance between the required detection defect performance, the time allowed for calculation time, and the time required for image processing for defect extraction. Actually, since the form of defects increases, a margin for absorbing these in the future may be provided. Further, the fact that each condition setting and imaging condition corresponding to such a newly generated defect can be additionally changed is often the point that determines whether or not the apparatus can be used in the field.

(Step S6: Detection of abnormal part)
Next, the abnormality detection unit 87 uses the abnormal part extraction filter and its optimum parameters from the developed image data generated by the image reconstruction unit 83 based on the condition data stored in the condition information storage unit 101 to detect an abnormal part ( (Abnormal area) is detected.

  Further, the abnormality detection unit 87 detects an abnormal part from a plurality of developed image data by detecting each unique abnormal part from the correlation of the plurality of developed image data in which the same positions of the plurality of developed image data are matched by the correcting unit 84. Processing can be performed with the optimum parameters. This process is performed on the developed image data under the condition that the developed image data under one condition cannot secure a sufficient S / N and may be erroneously recognized, or the developed image data under the condition that only a part of the abnormal part can be recognized. It will be used against.

(Step S7: Determination of abnormal part)
Next, the abnormality determination part 88 determines the abnormal part detected by the abnormality detection part 87 based on a predetermined condition, and outputs a determination result. Specifically, the determination performed by the abnormality determination unit 88 is a determination based on a determination criterion of a level required for the appearance inspection, and there are the following two methods for the determination.

  First, there is a method of detecting other than normal parts as abnormal parts, displaying them as necessary, identifying and classifying the abnormal parts as defective parts and others, rejecting defective products, and determining others as suspended products is there. This determination method can ensure a normal portion of the cylindrical surface within a certain range, and can select rejected products while monitoring the abnormal state. Moreover, it can also classify | categorize finely according to the kind and grade of abnormality as needed.

  In addition, as another determination method, there is a method of determining a defect sample or a defect that is worse than a defined defect as a failure. This method is a method corresponding to a screening inspection method used when supplying a product that maintains the quality within the range promised by the contract for the level of failure. Usually, a comparative evaluation is applied to a degree that is stricter by the difference and variation in the judgment capability between the person and the cylindrical body appearance inspection apparatus 10.

  Note that the conditions, methods, and the like when the abnormality determination unit 88 performs the above processing are based on the analysis results of standard samples such as normal parts and defects to be applied and the determination method, and the optimal abnormality part determination filter and its optimal parameters Is set in advance.

  Further, the developed image data generated by the image reconstruction unit 83 described above, the abnormal part detected by the abnormality detection unit 87, and the determination result by the abnormality determination unit 88 are stored in association with the image storage unit 100. In addition, the operator may operate the operation unit 82 to output the image monitor 82 via the output unit 85. Further, these data may be transmitted to another device via the communication unit 86. Also, by recording these data, the performance of the cylindrical body appearance inspection apparatus 10 is maintained, and whether or not an abnormality has occurred in the process of the cylindrical body inspection sample 1 that is a manufactured product. It becomes a device that can be monitored and the past progress can be investigated and analyzed.

  In general, the detected data is fed back and used for better process management and reviewing the validity of comparative evaluation.

  The pass / fail of these inspections and the hold result may also be transmitted by the communication unit 86, and the corresponding cylindrical body inspected sample 1 may be classified into pass, fail, or hold by the destination device.

  In step S4, when imaging for one round or more is completed, the roller drive motor 62 is stopped to stop the rotation of the roller 61 (step S3-1), and the cylindrical body is inspected. It is determined whether or not the imaging for all the inspection target areas of the sample 1 has been completed (step S3-2), and the imaging for all the inspection target areas of the cylindrical specimen 1 has not been completed. After the axial conveyance unit 9 lifts the cylindrical specimen 1 and moves it to the next inspection position (step S3-3), the processing from step S1 is repeated again. On the other hand, if the imaging for all the inspection target areas has been completed, the abnormal portions of all the areas are comprehensively determined, and the inspection process is ended.

  With the above configuration, illumination is performed so that a striped pattern that continuously changes in multiple values parallel to the axis of the cylindrical inspection sample 1 is reflected in the captured image, and the cylindrical inspection sample 1 is rotated once to make one camera. 5, it is possible to obtain a captured image in a large range from the bright field to the dark field including the intermediate brightness field and the imaging angle before and after, and each generated from the image data of the captured image It is possible to perform abnormal part detection and determination using developed image data having different conditions. In the above configuration, using image data of one pixel line is the same as inspecting with a conventional line sensor, but with a conventional line sensor, it is easy to detect one defect that is an inspection target. Arrange the lighting and line sensor. In this case, the inspection is often performed under bright field illumination. On the other hand, in the above configuration, a plurality of developed image data becomes developed image data having different characteristic imaging angles under illumination conditions from each of a characteristic bright field to a dark field. It is possible to improve the S / N ratio and make an automatic appearance inspection closer to the visual inspection.

  In addition, with the above-described configuration, the cylindrical body appearance inspection apparatus 10 captures a plurality of illumination angles of about ± 50 degrees and about ± 50 degrees by photographing the cylindrical specimen 1 for one round or more than one round. Reflection of the circumferential surface of the cylindrical specimen 1 at a plurality of imaging angles, a plurality of illumination imaging angles of about ± 100 degrees, and a continuous plurality of brightnesses from the normal reflection bright field part to the dark field part. The developed image data by light can be obtained. Then, by correcting the deviation of each developed image data depending on the imaging position, it is possible to obtain developed image data corresponding to the position of the circumferential surface of the cylindrical specimen 1 on a one-to-one basis and having different imaging conditions. .

  Therefore, much more information on the cylindrical surface can be obtained than information on the cylindrical surface obtained by the line sensor. Thereby, an abnormal part can be extracted from single defect and a plurality of developed image data with good S / N of each defect and each normal part, and S / N of inspection can be improved.

  Furthermore, since a plurality of matched developed image data with different conditions can be obtained, the abnormal part can be complemented, and the determination condition for the abnormal part can be increased. The performance can be improved, and as a result, the S / N of the inspection can be further improved.

  In addition, said embodiment is a standard structure, and the magnitude | size, arrangement | positioning location, space | interval, etc. of illumination are the size of the cylindrical body of the cylindrical body to-be-tested sample 1, a curvature, the state of a normal surrounding surface, and the defect which should be detected. It will be adjusted according to the required resolution of.

  Further, the number of the sample axis parallel illumination 71 or the sample axis parallel oblique illumination 72 can be arbitrarily determined. For example, one thick sample axis parallel illumination 71 is used as in the captured image of FIG. You may make it test | inspect.

  Further, as a general rule, the brightness of the illumination is adjusted so that a striped pattern substantially similar to a captured image obtained by capturing the normal portion of the cylindrical surface of the cylindrical specimen 1 with the camera 5 appears in the captured image. Depending on the surface condition of the cylindrical surface of the cylindrical specimen 1 and the defect to be detected preferentially, the brightness balance may be changed so that the defect can be easily detected.

  Further, the imaging lens of the camera 5 uses a lens having an angle of view as inspected with one eye of visual inspection. This is to inspect by imaging under conditions close to visual inspection. However, in inspections where importance is placed on the reproducibility of inspection, cylindrical objects in the center of the lens can be inspected by using a telecentric imaging lens for this imaging lens. Even if the axial position of the sample 1 changes, the reproducibility can be made not to change so much.

  If there is a margin, as inspecting with both eyes of the visual inspection, attach two cameras 5 to the cylindrical body appearance inspection apparatus 10 and send the captured images of these two cameras to the image processing apparatus 8, If developed images from the two cameras 5 are created and defects are extracted using a three-dimensional algorithm, an automatic appearance inspection apparatus closer to visual inspection is obtained.

  The appearance inspection apparatus described in the present invention corresponds to a cylindrical body appearance inspection apparatus. The image processing unit described in the present invention corresponds to the transmission / reception unit 81, the image storage unit 100, the image reconstruction unit 83, and the correction unit 84 in the image processing apparatus 8. The abnormality inspection unit described in the present invention corresponds to the abnormality detection unit 87 and the abnormality determination unit 88.

  The above-described image processing apparatus 8 has a computer system inside. The process of receiving and processing the image data captured by the camera 5 of the cylindrical specimen 1 described above is stored in a computer-readable recording medium in the form of a program, which is read by the computer. By executing this, the above processing is performed. Here, the computer-readable recording medium means a magnetic disk, a magneto-optical disk, a CD-ROM, a DVD-ROM, or the like. Alternatively, the computer program may be distributed to the computer via a communication line, and the computer that has received the distribution may execute the program.

It is a front view of the cylindrical body external appearance inspection apparatus which concerns on embodiment of this invention. It is a side view of the cylindrical body external appearance inspection apparatus which concerns on the same embodiment. It is a side view of the axial direction conveying machine concerning the embodiment. It is a schematic block diagram of the image processing apparatus which concerns on the same embodiment. It is the picked-up image (the 1) of the cylindrical body test sample imaged with the camera which concerns on the embodiment. It is the flowchart which showed the flow of the process of the cylindrical body external appearance inspection which concerns on the same embodiment. It is the picked-up image (the 2) of the cylindrical body test sample imaged with the camera which concerns on the same embodiment. It is a picked-up image (the 3) of the cylindrical body test sample imaged with the camera which concerns on the embodiment. It is the figure which showed the expansion | deployment image reconstructed from the captured image which concerns on the same embodiment. It is the image (the 4) which expanded a part of captured image of FIG. 8 for description.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Cylindrical body test sample 10 Cylindrical body external appearance inspection apparatus 5 Camera 51 Imaging lens 52 Two-dimensional image sensor 53 Light-shielding plate 6 Rotating part 61 Roller 62 Roller drive motor 7 Illumination part 71 Sample axis parallel illumination 72 Sample axis parallel oblique illumination 73 Light shielding plate 75 Illumination support 8 Image processing device 9 Axial transport unit 91 Drive motor 92 Axial drive roller 93 Drive roller lift-up function 200 Image monitor

Claims (2)

A visual inspection apparatus for inspecting a cylindrical surface of a cylindrical body inspection sample,
A rotating unit for rotating the cylindrical object sample;
An illuminating unit arranged so that a striped pattern parallel to the cylinder axis is reflected on a cylindrical surface of the cylindrical object to be inspected;
A camera that is arranged so that the cylindrical body axis and the optical axis are at right angles, and images the area of the cylindrical surface including the striped pattern at a speed corresponding to a predetermined circumferential resolution, and outputs image data; ,
An image processing unit that extracts predetermined line pixel data from image data of a plurality of images output from the camera, reconstructs the extracted line pixel data, and generates expanded image data;
Based on the developed image data generated by the image processing unit, an abnormal portion of the cylindrical surface of the cylindrical body inspection sample is detected, and the abnormality of the cylindrical surface of the cylindrical body inspection sample is evaluated by the detected abnormal portion. An abnormality inspection unit for determining pass / fail of an inspection sample ,
The image processing unit generates a plurality of the developed image data for each of the line pixel data, and generates a plurality of the developed image data in a line of extraction source pixel data so as to correspond to the position of the cylindrical surface. Correct the pixel position accordingly,
The abnormality inspection unit matches an identical position of a plurality of developed image data corrected by the image processing unit, and detects an abnormal part from the plurality of developed image data using an abnormal part extraction filter. Inspection device. The illumination unit is arranged on the left and right with the central axis of the camera sandwiched so that the striped pattern appears in the captured image of the cylindrical surface of the camera parallel to the cylindrical body axis of the cylindrical body inspection sample. The visual inspection apparatus according to claim 1, further comprising a parallel illumination of the sample axis.

Images