JP2008170256A - Defect detection method, defect detection program, and inspection apparatus - Google Patents

Abstract

A defect detection method, a defect detection program, and an inspection apparatus capable of accurately detecting streak defects due to grouping of point defects are provided.
An area in which point defect candidates are gathered (grouped) more than a certain degree in a captured image is determined as a target area (ST4: area determination step), and cluster defects in which point defect candidates are continuous are calculated (ST5). : Labeling step), calculating the linearity and separation distance between cluster defects and determining that the cluster defect is a streak defect candidate (ST6: defect classification step), then detecting the streak defect by the procedure for detecting streak defects By performing (ST8: defect detection step), it is possible to accurately detect streak defects due to grouping of point defects and improve the accuracy of defect determination.
[Selection] Figure 2

Description

  The present invention relates to a defect detection method, a defect detection program, and an inspection apparatus that detect a display defect of a display device based on a captured image acquired by capturing a display image of a display device to be inspected.

  In the inspection process in the manufacture of display devices such as liquid crystal panels and projectors that are applied products, CCDs are used to detect spot defects such as bright spots and dark spots, surface defects such as spots and unevenness, and stripe defects by image processing. A target image is acquired by a camera or the like, and image processing is performed on the acquired image to detect the image.

  As a method for detecting a bright spot defect or a dark spot defect from an image, a method using a filter for emphasizing a defect (an Tophat filter for emphasizing a bright spot defect or a Well filter for enhancing a dark spot defect) is used. It has been.

  On the other hand, as a method for detecting a streak defect (streaks defect) from an image, a method for detecting streak defects by performing integration processing in a specific direction on the image (for example, see Patent Document 1), or for an image For example, a method of voting candidate points far from the target pixel for each angle and detecting streak defects by integrating the vote scores (see, for example, Patent Document 2) has been proposed.

Japanese Patent Laid-Open No. 11-53557 JP-A-10-289319

  By the way, as a display defect of a display device, a defect (streaks due to grouping of point defects) may appear in which bright spots and dark spots are discontinuous and arranged in a substantially straight line (grouped). Such streak defects due to grouping of point defects are difficult to detect by the defect detection methods described in Patent Document 1 and Patent Document 2, and can be detected only by visual inspection by an inspector. First, it was a problem in terms of inspection efficiency.

  In other words, in the techniques of Patent Documents 1 and 2, it is not possible to recognize a streak defect unless the low-luminance (or high) portion is in a linear state, and the point defect component is eliminated by the noise removal process. Thus, it becomes impossible to detect the point defect itself, and it is also impossible to detect the streak defect by grouping the point defects.

  The present invention has been made in view of the above-described problems, and an object of the present invention is to provide a defect detection method, a defect detection program, and an inspection apparatus that can accurately detect streak defects due to grouping of point defects.

  The defect detection method of the present invention is a defect detection method for detecting a display defect of the display device based on a captured image obtained by capturing a display image of a display device to be inspected. A captured image acquisition step of capturing and acquiring a captured image, a point defect extraction step of filtering the captured image to extract point defect candidates in the captured image, and searching the captured image in a region of a predetermined size A region search step of detecting the number of point defect candidates included in the region, and a region determination step of determining whether or not the point defect candidates are included in a predetermined number or more for each of the regions searched in the region search step A labeling step of calculating and storing a cluster defect in which a number of point defect candidates equal to or greater than a predetermined threshold are related to a target region including a predetermined number or more of point defect candidates; A defect classification step for determining whether or not the defect is a streak defect candidate by calculating the linearity between the cluster defects calculated in the ring process and the separation distance between the cluster defects, and a streak defect candidate in the defect classification process A defect detection step of detecting a defect using a streak defect detection procedure.

  In the present invention, first, a point defect candidate (bright spot defect or dark spot defect) in a captured image is extracted in a point defect extraction step, and then the number of point defect candidates included in a region of a predetermined size is determined in a region search step. It is detected and it is determined in the region determination step that a predetermined number or more of point defect candidates are included. Through these series of processes, an area in which point defect candidates are gathered (grouped) more than a certain degree in a captured image can be determined as a target area.

  Next, after calculating the cluster defect in which the point defect candidates are continuous in the labeling process, the linearity and separation distance between the cluster defects are calculated in the defect classification process, and it is determined that the cluster defect is a streak defect candidate. In such a case, a streak defect detection procedure is used in the defect detection step. In other words, by treating grouped point defect candidates as streak defects and using a detection procedure for detecting this streak defect, it is possible to accurately detect streak defects due to grouping of point defects and improve defect determination accuracy. it can.

  Here, when the grouping state of the point defect candidates is not determined as a streak defect, that is, when the grouping state is rough or not positioned on a straight line, it is not a streak defect but individual as in the conventional case. It can be treated as a point defect.

  In the present invention, in the defect classification step, a center-of-gravity coordinate for each cluster defect is calculated, a Hough transform is executed based on the calculated center-of-gravity coordinate, and the clusters are obtained by overlapping of ρ-θ curves obtained by the Hough transform. It is preferable to calculate the linearity between defects and determine that the cluster defects are the same streak defect candidates when a predetermined number or more of the ρ-θ curves overlap each other.

  According to such a configuration, a ρ-θ curve relating to each barycentric coordinate is obtained by executing the Hough transform based on the barycentric coordinate calculated for each cluster defect. In this ρ-θ curve, ρ is the distance from an arbitrary origin of the target area to the center of gravity coordinates, and θ is a straight line connecting the origin of the target area and the center of gravity coordinates to an arbitrary reference axis (X axis, Y axis, etc. ). If such a ρ-θ curve is calculated for each barycentric coordinate and the ρ-θ curve is graphed, it is known that the ρ-θ curves of the barycentric coordinates located on the same straight line overlap at a certain point. ing. Therefore, when the ρ-θ curves related to the barycentric coordinates for each cluster defect overlap, it can be determined that the cluster defects are located on the same straight line.

  Note that the determination of whether or not the ρ-θ curves overlap is not limited to the case where the ρ-θ curves overlap at the same point, but also includes the case where the ρ value and the θ value overlap within a predetermined numerical range. What is necessary is just to be judged including the case where it is located within the linear range having a predetermined width.

  In the present invention, in the defect classification step, the separation distance is calculated based on the number of non-defective pixels between adjacent cluster defects. When the number of non-defective pixels is equal to or less than a predetermined number, the cluster defects are identified as the same stripe. It is preferable to determine as a defect candidate.

  According to such a configuration, the separation distance between the cluster defects is calculated based on the number of non-defective pixels between adjacent cluster defects, and the cluster defects having the number of non-defective pixels equal to or less than a predetermined number (small separation distance) are detected in the same stripe. By determining as a defect candidate, even if they are separated from each other, cluster defects positioned within a predetermined separation distance can be determined as one streak defect candidate. On the other hand, when the number of non-defective pixels between cluster defects is equal to or greater than a predetermined number (the separation distance is large), each cluster defect is determined as another streak defect candidate. Thus, it can be determined as one streak defect candidate according to the separation distance between cluster defects, or by determining as another streak defect candidate, the same determination as the visual defect determination can be performed, and display The display quality as a device can be ensured.

  In the present invention, it is preferable that in the labeling step, a binarization process is executed to separate the cluster defect and a portion other than the cluster candidate into a binary value.

  According to such a configuration, it is necessary to process a halftone pixel by performing binarization processing to separate a pixel having a cluster defect and other pixels and then performing the subsequent processing. It can be eliminated and the processing speed can be increased.

  The present invention further includes a feature amount calculation step of calculating a defect feature amount of the streak defect candidate determined in the defect classification step, wherein the defect feature amount includes an area value of the streak defect candidate and the streak defect candidate. It is preferable that at least the brightness average value of the above and the brightness slope value of the streak defect candidate are included.

  According to such a configuration, by calculating the area value, brightness average value, and brightness slope value of the streak defect candidate in the feature amount calculation process, the accuracy of streak defect detection in the subsequent defect detection process is improved, and The speed of the detection process can also be improved.

  On the other hand, the defect detection program of the present invention causes a computer to execute the defect detection method.

  According to this defect detection program, by using a computer equipped with a CPU, it is possible to increase the processing speed of each process and quickly execute the entire defect detection process.

  The inspection apparatus of the present invention is an inspection apparatus that detects a display defect of the display device based on a captured image acquired by capturing a display image of a display device to be inspected, and displays the display image of the display device. An imaging device that captures an image; and a control device that drives and controls the display device and the imaging device. The control device causes the imaging device to capture a display image of the display device, acquires the captured image, and the imaging Filter the image to extract point defect candidates in the captured image, search the captured image in an area of a predetermined size, detect the number of point defect candidates included in the area, and for each searched area Whether or not the number of point defect candidates is greater than or equal to a predetermined number and the number of point defect candidates equal to or greater than a predetermined threshold with respect to a target region including the predetermined number of point defect candidates When calculating and storing continuous cluster defects, calculating the linearity and separation distance between the cluster defects to determine whether or not it is a streak defect candidate, and when determining that it is a streak defect candidate The defect detection is performed using a streak defect detection procedure.

  According to the inspection apparatus of the present invention, as in the above-described defect detection method, an area where point defect candidates are grouped in a captured image is determined as a target area, and a cluster defect is determined based on the grouping state of point defect candidates. By calculating the linearity and separation distance between cluster defects, the cluster defect is treated as a streak defect when it is determined to be a streak defect candidate, and the detection procedure for detecting this streak defect is By using this, it is possible to accurately detect streak defects due to grouping of point defects and improve the accuracy of defect determination.

Embodiments of the present invention will be described with reference to the drawings.
[Configuration of inspection equipment]
FIG. 1 is a diagram illustrating a configuration of the inspection apparatus 1.

  The inspection apparatus 1 is an apparatus that performs a display appearance inspection of the liquid crystal panel 10 as an image display device to be inspected and detects a display defect of the liquid crystal panel 10. As shown in FIG. 1, the inspection apparatus 1 includes a light source 2 that emits a light beam on a liquid crystal panel 10, a lens 3 that enlarges and projects a light beam that passes through the liquid crystal panel 10, and a light beam that passes through the lens 3 on a screen 5. Mirror 4 for projecting, screen 5 provided in a dark box 6 for blocking external light, a CCD camera 7 as an imaging device, a pattern generator 8 for displaying an appropriate inspection pattern on a liquid crystal panel 10, and a control device And a computer 9.

  Here, the liquid crystal panel 10 to be inspected is a transmissive liquid crystal panel, and has, for example, a configuration in which liquid crystal molecules are hermetically sealed between a TFT substrate and a counter substrate. It has an image display part to be transmitted. For example, the liquid crystal panel 10 is electrically connected to the pattern generator 8 in a state where the liquid crystal panel 10 is installed in a panel installation unit (not shown) in the inspection apparatus 1. When the voltage value (including 0 V) is applied, the arrangement state of the liquid crystal molecules is changed, and a predetermined optical image is formed by transmitting or blocking the incident light beam. In the present embodiment, the liquid crystal panel 10 is configured in a normally white mode in which white display is performed by transmitting all incident light beams in a state where no voltage is applied (voltage value is 0 V). Further, the liquid crystal panel 10 may be configured in a normally black mode in which a black display is performed by blocking an incident light beam when no voltage is applied.

  Examples of the light source 2 include a discharge light source lamp, a light emitting diode (LED) element, a laser diode, an organic EL (Electro Luminescence) element, a silicon light emitting element, and other various solid light emitting elements. An appropriate light source can be employed.

  The screen 5 is configured as a reflective screen that reflects and projects an optical image (displayed image) enlarged and projected by the lens 3 and the mirror 4. The screen 3 is not limited to a reflective screen, and may be configured as a transmissive screen that transmits and projects an incident optical image.

  The CCD camera 7 images the projection surface of the screen 5 under the control of the computer 9 and outputs a signal corresponding to the captured image to the computer 9. Although not specifically shown, the CCD camera 7 receives the incident light beam by the CCD under the control of the CCD, which is an area sensor, a condensing lens that collects the light beam and irradiates the CCD, and the computer 9. It has a shutter that can change the time.

  The CCD of the CCD camera 7 preferably has a resolution higher than that of the liquid crystal panel 10.

  The pattern generator 8 applies a voltage having a predetermined voltage value to the liquid crystal panel 10 under the control of the computer 9 to form a predetermined optical image on the liquid crystal panel 10.

  The computer 9 includes, for example, a CPU (Central Processing Unit) that reads and executes a predetermined program, and controls the entire inspection apparatus 1. The computer 9 includes a control unit that executes predetermined processing (defect detection processing) according to a control program, and a memory that stores the control program and various data.

  Although not shown, the control unit includes an image display control unit that controls the pattern generator 8, an image data acquisition unit that controls the CCD camera 7, an image processing unit that processes the acquired image, and the liquid crystal panel 10 based on the image. And a defect detection unit for detecting the display defect.

  The image display control unit reads voltage value information related to the voltage value applied to the liquid crystal panel 10 stored in the memory, and patterns a predetermined control signal for driving the liquid crystal panel 10 with the test voltage value based on the voltage value information. Output to the generator 8. Then, the pattern generator 8 drives the liquid crystal panel 10 with the inspection voltage value according to the control signal output from the image display control unit, and causes the liquid crystal panel 10 to form an inspection image.

  The image data acquisition unit outputs a predetermined control signal to the CCD camera 7 to cause the CCD camera 7 to image the projection surface of the screen 5 when an inspection image is formed on the liquid crystal panel 10. The image data acquisition unit receives an electrical signal output from the CCD camera 7 and converts it into a computer-readable signal (digital signal), and includes information on a pixel value (luminance value) for each pixel. Inspection image data (captured image data) P is acquired, and the acquired inspection image data P is stored in a memory.

  The image processing unit executes various types of image processing such as performing shading correction on the acquired inspection image data P and performing image processing such as contraction processing and expansion processing of the inspection image data P.

  The defect detection unit detects a defect in the inspection image data P by executing a defect detection procedure described below. That is, a point defect is enhanced by applying various filters to the inspection image data P, an area including a predetermined number of point defects is detected by searching the inspection image data P, and an object including a predetermined number or more of point defect candidates Calculate the grouping state of point defect candidates for the region, determine whether or not it is a streak defect based on the calculated grouping state, and if it is determined that it is a streak defect, use the streak defect detection procedure Detect defects.

The memory stores various filter values, calculation formulas, and thresholds in addition to a predetermined control program, information for executing processing by the control unit, data processed by the control unit, and the like.
(Defect detection method)
Next, a defect detection method using the above-described inspection apparatus 1 will be described with reference to the drawings.

  FIG. 2 is a flowchart for explaining the defect detection method.

  In the following, it is assumed that the liquid crystal panel 10 is installed in a panel installation unit (not shown) in the inspection apparatus 1 and the liquid crystal panel 10, the pattern generator 8, and the computer 9 are electrically connected.

  The operator operates an operation unit (not shown) of the computer 9 to perform setting input for executing defect detection of the liquid crystal panel 10. Then, the control unit of the computer 9 reads out the control program stored in the memory in accordance with the operation signal output from the operation unit, and performs defect detection processing of the liquid crystal panel 10 according to the control program as shown below. Execute.

  First, the control unit of the computer 9 acquires an inspection image and executes preprocessing for performing shading correction on the acquired inspection image data P (processing ST1: preprocessing step).

  Specifically, the control unit outputs a command to a receiving unit (not shown) provided on the light source 2 side to turn on the light source 2 and outputs a command to the pattern generator 8 to display an inspection image on the liquid crystal panel 10. Display. By this process, the inspection image transmitted through the liquid crystal panel 10 is projected and displayed on the screen 5. In this state, the image data acquisition unit of the computer 9 outputs a predetermined control signal to the CCD camera 7, causes the CCD camera 7 to image the projection surface of the screen 5, acquires the inspection image data P, and acquires the acquired inspection image data. P is stored in the memory (captured image acquisition step). Although the lighting of the light source 2 is controlled by the control unit of the computer 9 here, the light source 2 is not limited to the lighting controlled by the control unit, but can be switched on and off by itself. May be adopted.

  Next, the image processing unit of the computer 9 reads the acquired inspection image data P and the background image data prepared and stored in the memory in advance, and for each corresponding pixel of the background image data from the inspection image data P. A process for obtaining a difference between luminance values is performed (shading correction). Then, the image processing unit stores the inspection image data P subjected to the shading correction in the memory.

  After the processing ST1 as described above, the defect detection unit of the computer 9 extracts a point defect (bright spot or dark spot) component in the image from the inspection image data P subjected to the shading correction, and a bright spot / dark spot enhancement process. (Process ST2: Bright spot / dark spot enhancement process, point defect extraction process). Here, the coordinates and luminance values of the points highlighted with the bright spot / dark spot enhancement processing and the coordinates and luminance values of the points highlighted with the dark spot are stored in the memory as separate arrays, and processing ST3 is performed. Thereafter, each process is performed on each of the inspection image data P as an array with bright spot emphasis and the inspection image data P as an array with dark spot emphasis.

  The bright spot / dark spot enhancement process will be described in detail. With each pixel in the inspection image data P as a target pixel, a filter having a predetermined size (for example, a 3 × 3 matrix) around the target pixel is applied. It is detected whether the target pixel is larger or smaller than a predetermined ratio (for example, 20%) than the average luminance value of the surrounding pixels (in the case of 3 × 3, the surrounding 8 pixels).

  That is, if the luminance value of the target pixel is larger than a predetermined ratio (for example, 1.2 times) or more with respect to the average value of the luminance values of surrounding pixels (maximum value), the target pixel is stored as a bright spot (point defect candidate). Store in the array. On the other hand, if the luminance value of the target pixel is smaller than a predetermined ratio (for example, 0.8 times) or less with respect to the average value of the luminance values of the peripheral pixels (minimum value), the target pixel is stored as a dark spot (point defect candidate). Store in the array. Further, the luminance value of the target pixel that has not been detected as the maximum value or the minimum value is not stored in the array, that is, only the pixels detected as the maximum value or the minimum value are emphasized and stored in the array. Point defect candidates that are bright spots or dark spots are emphasized.

  After the processing ST2, the defect detection unit of the computer 9 searches the inspection image data P in which the point defect candidates are emphasized in a predetermined size area and detects the number of point defect candidates included in the area. A search is performed (process ST3: candidate area search step, region search step).

  Specifically, as shown in FIG. 3, any (for example, 5 × 5 pixels to 31 × 31 pixels) depending on the size of the grouped point defects to be detected (streaks due to the grouping of point defects). Using the search area Q as the region set in (2), the inspection image data P is searched, and the number of point defect candidates included in the search area Q is detected. Then, the coordinate position of the search area Q and the number of detected point defect candidates are stored in the memory.

  After process ST3, the defect detection unit of the computer 9 determines whether or not a predetermined number or more of point defect candidates are included (presence / absence of candidates) for each search area Q at each coordinate position stored in the memory (process). ST4: Area determination step).

  Specifically, a predetermined ratio (for example, 1.5% to 30%) with respect to the number of pixels in the search area Q (for example, 25 for 5 × 5 pixels and 961 for 31 × 31 pixels). %)), It is determined whether there are more point defect candidates than the number of pixels. When the number of point defect candidates is greater than or equal to a predetermined ratio with respect to the number of pixels in the search area Q (YES in process ST4, with candidates), the following processes ST5 to ST8 are executed. On the other hand, if the number of point defect candidates is less than the predetermined ratio with respect to the number of pixels in the search area Q (NO in process ST4, no candidate), it is assumed that there are no grouped point defects in the search area Q. The detection process ends.

  In the process ST5, as shown in FIGS. 4 and 5, the defect detection unit of the computer 9 has a target area of a predetermined size including the search area Q determined to include a predetermined number or more of point defect candidates in the process ST4. A binarization process is performed on the voting area (R), and cluster defects d1 to d8 in which a number of point defect candidates equal to or greater than a predetermined threshold are consecutively calculated and stored (process ST5: labeling step).

  Specifically, the same threshold value as the threshold value for detecting defect candidates in process ST4 is used, and when the number of point defect candidates equal to or greater than this threshold value is consecutive (cluster defect), 1 (white) is set, and other values are set to 0. The binarization process is executed so that (black). In FIGS. 4 and 5, white and black are reversed. After this binarization processing, the defect detection unit prepares an array having the same size as the target region R in the memory, and stores the binarized values in this array.

  Next, as shown in FIG. 5, the defect detection unit attaches identification numbers (1 to 3 in FIG. 5) to the cluster defects d1 to d3 in which the point defect candidates are continuous, and has the same size as the target region R. An array having a length is prepared in the memory, and an identification number is stored in this array.

  After the process ST5, the defect detection unit of the computer 9 calculates the linearity between the cluster defects d1 to d8 and the separation distance between the cluster defects d1 to d8 calculated in the process ST5, and determines whether or not the defect is a streak defect candidate. (Process ST6: Defect classification step).

  Specifically, first, as shown in FIG. 5, the defect detection unit calculates the barycentric coordinates (positions indicated by white crosses in FIG. 5) for each of the cluster defects d1 to d8 assigned with the identification numbers. Next, based on the calculated barycentric coordinates, the defect detection unit performs a Hough transform, and calculates the linearity between the cluster defects d1 to d8 based on the overlap of the ρ-θ curves obtained by the Hough transform. In this case, the ρ-θ curve obtained by the Hough transform is that ρ is the distance from an arbitrary origin of the target region R to the center of gravity coordinates of the individual cluster defects d1 to d8, and θ is the origin and the center of gravity of the target region R. The straight line connecting the coordinates is an angle formed with an arbitrary reference axis (for example, the X axis), and has the relationship of the following expression (1).

ρ = xi · cosθ + yi · sinθ (1)
Here, (xi, yi) is the barycentric coordinates of the cluster defects d1 to d8.

  The above ρ-θ curve can be graphed as shown in FIG. 6. In this graph, when the ρ-θ curves of the cluster defects d1 to d8 overlap each other, the corresponding cluster defects are the same. It is known to lie on a straight line. Therefore, the defect detection unit detects a point (near θ = 25 ° in FIG. 6) where the ρ-θ curves of the cluster defects d1 to d8 overlap each other by a predetermined number (for example, 5) or more, and this overlapping point exists. In this case, it is determined that the corresponding cluster defects d1 to d8 are located on the same straight line, and that these cluster defects d1 to d8 are streak defect candidates.

  Here, the determination as to whether or not the ρ-θ curves overlap includes not only the case where the ρ-θ curves overlap at the same point, but also the case where the ρ value and the θ value overlap within a predetermined numerical range.

  Next, the defect detection unit calculates a separation distance between adjacent cluster defects d1 to d8 in the streak defect candidate, and determines that each is a different streak defect candidate if the separation distance is equal to or greater than a predetermined threshold. . Specifically, as shown in FIG. 5, the separation distance is calculated based on the number of non-cluster defect pixels located between the cluster defects d1 to d8, and the number of non-cluster defect pixels is equal to or less than a predetermined number (for example, 4 In the case of (within square), adjacent cluster defects d1 to d8 are determined as the same streak defect candidate. On the other hand, when the number of non-cluster defect pixels exceeds a predetermined number, it is determined as another streak defect candidate. That is, in FIG. 4, the cluster defects d1 to d3 are determined to be the same streak defect candidates, the cluster defects d4 to d5 are determined to be the same streak defect candidates, and the cluster defects d6 to d8 are determined to be the same streak defect candidates. It has become so.

  In the above process ST6, when the defect detection unit determines that a streak defect candidate exists in the target region R (“streaks” in process ST6), the defect detection unit performs the following processes ST7 to ST8. On the other hand, if it is determined that there is no streak defect candidate in the target region R (“other than that” in process ST6), the target region R is assumed to have no streak defect in which point defects are grouped, and the defect detection processing is performed. Exit.

  In the process ST7, the defect detection unit of the computer 9 calculates the defect feature amount of the streak defect candidates (cluster defects d1 to d3, d4 to d5, d6 to d8) determined to be the same in the process ST6 (process ST7: feature). Amount calculation step).

  Here, the defect feature amount includes an area value of the same streak defect candidate, a luminance average value, a luminance inclination value, and the like. The area value is the total area (area) of the cluster defects d1 to d3, d4 to d5, d6 to d8 in the same streak defect candidate, and the luminance average value is the cluster defects d1 to d3, d4 to d5, d6 to d8. The brightness slope value is obtained by calculating the brightness maximum value of cluster defects d1 to d3, d4 to d5, and d6 to d8 in the same streak defect candidate as 1 / of a short side direction distance W to be described later. It is a value divided by 2 (W / 2).

  And in order to calculate the area value and brightness | luminance inclination value of a streak defect candidate, as shown in FIG. 7, the defect detection part is a short side of the rectangle circumscribing the cluster defects d1-d3, d4-d5, d6-d8 The direction distance W is calculated. The short-side direction distance W is calculated from the individual centroids of the cluster defects d1 to d8 calculated in the process ST6 and the inertia principal axis connecting the centroids, and the cluster defects d1 to d3, d4 to It is obtained by calculating a rectangle circumscribing d5, d6 to d8.

  After calculating the defect feature amount of the streak defect candidate in the process ST7, the defect detection unit of the computer 9 detects the defect using the procedure for detecting the streak defect (process ST8: a streak determination process, a defect detection process).

  Specifically, based on the defect feature amount of the line defect candidate, it is determined whether or not the line defect candidate is a line defect. That is, a streak defect candidate is regarded as a streak defect, and it is detected as a streak defect (a streak defect due to a group of point defects) by determining whether the threshold value for detecting a streak defect is higher or lower than the threshold value. .

  The embodiment described above has the following effects.

  In the present embodiment, the point defect candidates in the inspection image data P are emphasized and easily detected in the bright spot / dark spot emphasizing process of process ST2, and then the search area is searched in the candidate area searching process of process ST3. By detecting the number of point defect candidates in Q and determining whether or not the search area Q is the target region in the region determination step of process ST4, point defect candidates gathered in the inspection image data P to a certain extent. A region (grouped) can be determined as a target region.

  Further, after calculating the cluster defects d1 to d8 in which the point defect candidates are continuous in the labeling process of the process ST5, the linearity and the separation distance between the cluster defects d1 to d8 are calculated in the defect classification process of the process ST6. It is determined whether or not it is a streak defect candidate. When the cluster defects d1 to d8 are streak defect candidates, the defect feature amount is calculated in the feature amount calculation step in process ST7, and then the threshold for detecting the streak defect is used in the defect detection step in process ST8. Thus, streak defects due to grouping of point defects can be detected with high accuracy, and the accuracy of defect determination can be improved.

  Then, by calculating the linearity between the cluster defects d1 to d8 by the Hough transform, and determining whether or not it is a streak defect based on this, even if each cluster defect is located away from each other, It is possible to easily and quickly calculate the linearity and execute the streak defect candidate determination. Furthermore, by determining whether the cluster defects d1 to d8 that are adjacent to each other apart from each other are one streak defect candidate, it is possible to perform the same determination as the visual defect determination, and display quality as a display device Can be secured.

  It should be noted that the present invention is not limited to the above-described embodiments, and modifications, improvements, and the like within the scope that can achieve the object of the present invention are included in the present invention.

  In the above-described embodiment, the inspection apparatus 1 that performs the display appearance inspection of the liquid crystal panel 10 as an image display device has been described. However, the image display device of the present invention is not limited to a liquid crystal panel that transmits light from a light source, but an organic EL ( Various self-luminous image display devices such as an electro luminescence panel and a plasma panel may be used.

  In the embodiment, the linearity between the cluster defects d1 to d8 is calculated by the Hough transform in the defect classification step (process ST6). However, the linearity calculation method is not limited to the Hough transform, and an appropriate filter process is performed. Etc. may be executed.

  In the defect classification step (process ST6), the method for calculating the separation distance between the cluster defects d1 to d8 is not limited to the determination based on the number of pixels of the non-cluster defects, and the distance between the centers of gravity of the cluster defects d1 to d8 is used. It is also possible to use other appropriate methods.

  Moreover, in the said embodiment, although the binarization process was performed with respect to the object area | region R in the labeling process (process ST5), the binarization process is not essential and may be abbreviate | omitted.

  Although the best configuration for carrying out the present invention has been disclosed in the above description, the present invention is not limited to this. That is, the invention has been illustrated and described primarily with respect to particular embodiments, but it is not intended to depart from the technical concept and scope of the invention. Various modifications can be made by those skilled in the art in terms of materials, quantity, and other detailed configurations.

  Therefore, the description limited to the shape, material, etc. disclosed above is an example for easy understanding of the present invention, and does not limit the present invention. The description by the name of the member which remove | excluded the limitation of one part or all of such restrictions is included in this invention.

The figure which shows the structure of the test | inspection apparatus of this invention. The flowchart explaining the defect detection method in the said inspection apparatus. The figure explaining the search method of the captured image acquired by the imaging means. The figure which shows the cluster defect in the object area | region in a captured image. The figure which expands and shows the cluster defect in the said object area | region. The graph which shows (rho) -theta curve used for the linearity calculation of the said cluster defect. The figure explaining the defect feature-value calculation method of a stripe defect candidate.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... Inspection apparatus, 7 ... CCD camera (imaging apparatus), 9 ... Computer (control apparatus), 10 ... Liquid crystal panel (display device), d1-d8 ... Cluster defect, P ... Inspection image data (captured image), Q ... Search area, R ... target area, ST2 ... point defect extraction process, ST3 ... area search process, ST4 ... area determination process, ST5 ... labeling process, ST6 ... defect classification process, ST7 ... feature amount calculation process, ST8 ... defect detection process .

Claims (7)

A defect detection method for detecting a display defect of the display device based on a captured image acquired by capturing a display image of a display device to be inspected,
A captured image acquisition step of acquiring a captured image by capturing a display image of the display device;
A point defect extraction step of filtering the captured image and extracting point defect candidates in the captured image;
An area search step of searching the captured image in an area of a predetermined size and detecting the number of point defect candidates included in the area;
A region determination step of determining whether or not the predetermined number of point defect candidates is included for each of the regions searched in the region search step;
A labeling step of calculating and storing a cluster defect in which a number of point defect candidates equal to or greater than a predetermined threshold are related to a target region including a predetermined number or more of the point defect candidates;
A defect classification step for determining whether or not the defect is a streak defect by calculating the linearity between the cluster defects calculated in the labeling step and the separation distance between the cluster defects;
A defect detection method comprising: a defect detection step of detecting a defect using a streak defect detection procedure when it is determined that the defect classification step is a streak defect candidate. The defect detection method according to claim 1,
In the defect classification step, a center-of-gravity coordinate for each cluster defect is calculated, and a Hough transform is performed based on the calculated center-of-gravity coordinate. A defect detection method characterized in that, when a predetermined number or more of the ρ-θ curves overlap each other, the cluster defects are determined as the same streak defect candidates. In the defect detection method according to claim 1 or 2,
In the defect classification step, the separation distance is calculated based on the number of non-defective pixels between adjacent cluster defects, and when the number of non-defective pixels is equal to or less than a predetermined number, the cluster defects are determined as the same streak defect candidate. A defect detection method characterized by: In the defect detection method in any one of Claims 1-3,
In the labeling step, a binarization process is executed to separate the cluster defect and a portion other than the cluster candidate into binary values. In the defect detection method in any one of Claims 1-4,
A feature amount calculating step of calculating a defect feature amount of a streak defect candidate determined in the defect classification step;
The defect detection method characterized in that the defect feature amount includes at least an area value of the streak defect candidate, a luminance average value of the streak defect candidate, and a luminance inclination value of the streak defect candidate.   A defect detection program for causing a computer to execute the defect detection method according to claim 1. An inspection apparatus that detects display defects of the display device based on a captured image acquired by capturing a display image of a display device to be inspected,
An imaging device for imaging a display image of the display device;
A control device for driving and controlling the display device and the imaging device;
The controller is
Causing the imaging device to capture a display image of the display device to obtain a captured image;
Filtering the captured image to extract point defect candidates in the captured image,
Search the captured image in a region of a predetermined size to detect the number of point defect candidates included in the region,
A cluster in which whether or not a predetermined number or more of the point defect candidates are included for each searched area is determined, and a number of point defect candidates equal to or greater than a predetermined threshold are consecutive with respect to a target area including the predetermined or more point defect candidates. Calculate and store defects,
The straight line and the separation distance between the cluster defects are calculated to determine whether or not the defect is a streak defect candidate, and when it is determined to be a streak defect candidate, a defect is detected using a streak defect detection procedure. Inspection device characterized by

Images