Feedback driving chip defect classification method and system based on visual analysis

Abstract

The invention discloses a feedback driving chip defect classification method and system based on visual analysis, and relates to the technical field of image processing. Comprising the following steps: establishing a database, collecting historical defect pictures, and carrying out label classification according to different defects; obtaining a to-be-detected image, converting the to-be-detected image into a single-channel detection pixel image, carrying out region division, and calculating a block constant of an error-prone region; carrying out adaptive histogram equalization in a normal region by adopting blocks with different sizes, and carrying out enhanced histogram equalization in an error-prone region according to block constants to obtain a detection image; and judging defect types according to the similarity, and carrying out defect classification. According to the method, region division is carried out on the single-channel detection pixel image, blocks with different sizes are adopted, adaptive histogram equalization is carried out in a normal region, enhanced histogram equalization is carried out in an error-prone region, the detection image is obtained, small local features in the detection image are emphatically reflected, the detection effect is improved, the operation complexity is reduced, and the calculation amount is reduced.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a feedback drive chip defect classification method and system based on visual analysis. Background Art

[0002] Feedback driver chip is an integrated circuit used to control and regulate signals or power in electronic devices. It monitors the output signal in real time and adjusts the input signal based on the feedback information to ensure system stability and accuracy. Such chips are widely used in motor control, power management, audio amplification and other fields. Feedback driver chip defect classification plays an important role in semiconductor manufacturing and quality control, and can effectively improve the quality and efficiency of chip production.

[0003] Feedback drive chips are generally classified by visual inspection, such as an intelligent chip surface defect detection method and system with patent publication number CN116523869A, which includes: extracting the first defect type from the preset chip defect types; forming a target chip set; extracting the first chip and obtaining the first chip feature, the first chip feature and the first defect type have a corresponding relationship; constructing a chip defect database and storing it in a defect detection support vector machine; collecting the target chip video of the target chip, obtaining the target chip feature set, and inputting it into the defect detection support vector machine to obtain the output result; obtaining the target detection result. The problem of long detection time, poor real-time performance and low test accuracy in the existing chip surface defect detection by manual visual inspection is solved, and the effect of improving the real-time, high efficiency and accuracy of chip surface defect detection is achieved.

[0004] Chip defect detection generally uses an optical imaging system to collect images of the chip, and then analyzes the images through image processing algorithms to detect defects on the chip surface, such as missing pins, short circuits, bridges, etc. This method has a fast detection speed and can detect some tiny defects that are difficult to detect by manual visual inspection. The above method realizes defect detection by preprocessing the image to be detected and performing feature recognition. However, in the detection of feedback drive chips, the defect location is generally small, and on the same production line, the location where defects are prone to occur is relatively fixed. Preprocessing the entire image cannot focus on detecting the location where defects are prone to occur, and the amount of calculation is large. Summary of the invention

[0005] The purpose of the present invention is to provide a feedback driver chip defect classification method and system based on visual analysis to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solution: a feedback driver chip defect classification method based on visual analysis, comprising: Establish a database, collect historical defect images and record the time when the historical defect images were generated, convert the historical defect images into single-channel historical pixel images, and classify them according to the different labels of the defects; Obtain the image to be detected, convert the image to be detected into a single-channel detection pixel map, divide the single-channel detection pixel map into normal areas and error-prone areas by using the regional prediction method according to the historical defect images, and calculate the block constant of the error-prone area by using the block constant calculation method; Adopting blocks of different sizes, performing adaptive histogram equalization in the normal area, and performing enhanced histogram equalization in the error-prone area according to the block constant, the detection image is obtained, which can improve the contrast of the single-channel pixel image and focus on reflecting the smaller local features inside the error-prone area, thereby improving the detection effect and reducing the complexity of calculation; The inspection image is compared with the single-channel historical pixel image, and the defect type is determined based on the similarity to classify the defects. When there is a defect, the inspection image is saved in the database to update the single-channel historical pixel image.

[0007] Preferably, the regional prediction method comprises: Convert the historical defect map into a single-channel historical pixel map and mark the defect locations; Reassign the channel value of the defect position and change the channel value of the remaining positions to 0, obtain the initial area map, and assign weights to the initial area map according to the generation time of the single-channel historical pixel map; According to the weights of the initial area maps, the initial area maps are added in a matrix manner to obtain the predicted area map, specifically: ,in To predict the channel value of the area graph at the coordinate (x, y), is the channel value of the i-th initial area graph at the coordinate (x, y), is the weight of the i-th initial region map, which is set by the technicians. Its value is related to the generation time of its corresponding single-channel historical pixel map. The earlier the time, the smaller the value. is the total number of single-channel pixel images; In the forecast area diagram, ≥ The pixels are marked, and the area composed of all marked pixels is the error-prone area. The area division threshold is set by the technical staff and represents the size of the divided area.

[0008] Preferably, the regional prediction method comprises: Convert the historical defect map into a single-channel historical pixel map, mark the defect location, and obtain the center pixel coordinates of the marked area; According to the central pixel coordinates of multiple defect positions, the coordinate relationship of the defect positions is established, specifically: ,in is the correlation coefficient between the horizontal and vertical coordinates of the center pixel coordinates, that is, the correlation coefficient between the x-coordinate and the y-coordinate. is the average value of the horizontal coordinates of each central pixel coordinate, is the horizontal coordinate value of the jth defect in the i-th single-channel historical pixel image, is the number of middle pixel coordinates of the i-th single-channel historical pixel image, is the ordinate value of the jth defect in the i-th single-channel historical pixel image, is the average value of the ordinate in each central pixel coordinate, is the total number of single-channel pixel images; According to the correlation coefficient between the horizontal and vertical coordinates of the central pixel coordinates, a distribution prediction model is established to calculate the defect constant of each pixel position in the image to be tested, specifically: ,in is the defect constant of the pixel with coordinates (x, y) in the image to be detected, is the standard deviation of the horizontal coordinates of each central pixel coordinate, is the standard deviation of the ordinates in each central pixel coordinate, is the correlation coefficient between the horizontal and vertical coordinates of the center pixel coordinates, is the average value of the horizontal coordinates of each central pixel coordinate, is the average value of the ordinate in each central pixel coordinate, It is the abbreviation of exponential function, which is used to represent the exponential function with the natural constant e as the base. ; The distribution prediction model is used to calculate the coordinates of each pixel in the image to be detected. ≥ The pixels are marked, and the area composed of all marked pixels is the error-prone area. It is the set area division threshold, which represents the area size of the divided area.

[0009] Preferably, the region division threshold can be updated according to a threshold iteration method, and the threshold iteration method includes: Check whether the defect position of the image to be inspected exceeds the error-prone area; When the defect position exceeds the error-prone area, detect whether the defect position exceeds the error-prone area in the positive and negative directions of the horizontal axis or the positive and negative directions of the vertical axis at the same time. When it exceeds the error-prone area at the same time, obtain the maximum distance of the excess and calculate the change in the area division threshold, specifically: ,in The threshold for setting the area division. The area division threshold calculated last time, is the maximum distance beyond the error prone zone, It is the update coefficient, which is set by the technicians and its value represents the sensitivity of the update amplitude; When the defect position does not exceed the error-prone area, the minimum distance between the edge of the defect position and the edge of the error-prone area is obtained, and the change in area division is calculated, specifically: ,in Indicates the number of times the area division threshold has been increased. The threshold for setting the area division. The area division threshold calculated last time, is the minimum distance between the edge of the defect position and the edge of the error-prone area, It is the update coefficient, which is set by the technician and its value represents the sensitivity of the update amplitude.

[0010] Preferably, the block constant calculation method includes: Record the number of pixels whose calculated values ​​exceed the maximum number of channels in the last enhanced histogram equalization calculation; Get the total number of pixels in the error-prone area detected last time, and calculate the block constant, specifically: ,in is the block constant, is the block constant calculated last time, It is the original channel value whose channel value is greater than the maximum channel value after enhanced histogram equalization calculation in the previous calculation. is the upper limit of the channel value of the single-channel detection pixel map, is the number of pixels with channel value i in the error-prone area, Calculate the coefficient for the block, set by the technician, is the total number of pixels in the error-prone area detected last time.

[0011] Preferably, the adaptive histogram equalization comprises: Divide the pixels in the normal area of ​​the single-channel detection pixel map into several non-overlapping blocks; Calculate each block separately: , ,in is the number of pixels with channel value i in the block, For the cumulative function, that is, the number of pixels in the block with channel values ​​i and below, The upper limit of the channel value in the single-channel detection pixel map, is the mapping channel value of pixel i in the single-channel detection pixel map in the detection map, is the upper limit of the channel value of the single-channel detection pixel map, and are the minimum and maximum values ​​in the cumulative function respectively; The detection map corresponding to the normal area can be obtained by mapping all the channel values ​​of the normal area in the single-channel detection pixel map.

[0012] Preferably, the enhanced histogram equalization includes: Divide the pixels in the error-prone area in the single-channel detection pixel map into several non-overlapping blocks; Calculate each block separately: , ,in is the number of pixels with channel value i in the block, is a cumulative function, that is, the number of pixels with channel values ​​i and below in the block, The upper limit of the channel value in the single-channel detection pixel map, is the mapping channel value of pixel i in the single-channel detection pixel map in the detection map, is the upper limit of the channel value of the single-channel detection pixel map, and are the minimum and maximum values ​​in the cumulative function, respectively. is the block constant; The detection map corresponding to the error-prone area can be obtained by mapping all the channel values ​​of the error-prone area in the single-channel detection pixel map.

[0013] Preferably, the enhanced histogram equalization also includes an optimization method for block edges, specifically: Mark all block edge pixels, calculate the mapping channel value of the target pixel in the adjacent block, and perform weighted summation of the mapping channel value according to the number of adjacent pixels in the corresponding adjacent block, specifically: ,in is the mapping channel value of the adjacent block z (including the block where the target pixel is located) to the target pixel, Optimize values ​​for edge pixels, is the number of adjacent pixels of the target pixel in the adjacent block z (including the block where the target pixel is located), is the total number of adjacent pixels of the target pixel; After calculating the edge pixel optimization values ​​of all block edge pixels, the channel values ​​of the target pixels are reassigned using the edge pixel optimization values ​​to obtain the optimized detection image.

[0014] Compared with the prior art, the present invention has the following beneficial effects: The single-channel detection pixel map is divided into a normal area and an error-prone area, and the block constant of the error-prone area is calculated by the block constant calculation method. Blocks of different sizes are used to perform adaptive histogram equalization in the normal area, and enhanced histogram equalization is performed in the error-prone area according to the block constant to obtain a detection map. This can improve the contrast of the single-channel pixel map and focus on reflecting the smaller local features inside the error-prone area, thereby improving the detection effect and reducing the complexity of the operation and the amount of calculation.

[0015] At the same time, the regional prediction method is used to calculate the boundaries of the normal area and the error-prone area in real time. The division of the error-prone area is updated according to the results of each detection. In addition, the block constant calculation method is used to update the block constant of the error-prone area to ensure the accuracy of the division of the error-prone area and improve the pixel contrast of the error-prone area in the detection image, thereby improving the accuracy of the detection results.

[0016] In enhanced histogram equalization, by optimizing the block edges, the sense of separation between each block is reduced, the smoothness of the transition between blocks is improved, the image detail retention and edge processing effects are better, the original features of the image can be better maintained during the image processing process, and the accuracy of detection is further improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flow chart of the feedback driver chip defect classification method of the present invention; Figure 2 It is a schematic diagram of the flow chart of the regional prediction method (embodiment 1) of the present invention; Figure 3 It is a schematic diagram of the flow of the regional prediction method (Example 2) of the present invention; Figure 4 It is a flow chart of the threshold value changing method of the present invention; Figure 5 It is a schematic diagram of the process of adaptive histogram equalization of the present invention; Figure 6 It is a schematic diagram of the process of enhanced histogram equalization of the present invention. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] In the present application, for ease of understanding, the method steps used do not need to be executed in the order of the steps in this embodiment during actual operation. In other embodiments, these steps may be performed simultaneously or in a different order.

[0020] Embodiment 1: In the detection of feedback driver chips, the defect locations are generally small, and on the same production line, the locations where defects are prone to occur are relatively fixed. When the feedback driver chips are defect-classified by visual inspection, preprocessing the entire image cannot focus on detecting the locations where defects are prone to occur, and the amount of calculation is large. This embodiment can perform partition detection on the detection image, reduce the amount of calculation, and refine the detection of error-prone areas, focusing on reflecting the smaller local features inside them, thereby improving the detection effect.

[0021] like Figure 1-Figure 6 As shown, the present invention provides a technical solution: a feedback driver chip defect classification method based on visual analysis, comprising: Establish a database, collect historical defect images and record the time when the historical defect images were generated, convert the historical defect images into single-channel historical pixel images, and classify them according to the different labels of the defects; Obtain the image to be detected, convert the image to be detected into a single-channel detection pixel map, divide the single-channel detection pixel map into normal areas and error-prone areas by using the regional prediction method according to the historical defect images, and calculate the block constant of the error-prone area by using the block constant calculation method; It should be noted that there are many ways to convert the image to be detected into a single-channel detection pixel map, such as the maximum channel number method, the average channel number method and the weighted channel number method, etc., which are all existing technologies and will not be repeated here. Different conversion methods can be used according to different acquisition equipment and production environments.

[0022] Adopting blocks of different sizes, performing adaptive histogram equalization in the normal area, and performing enhanced histogram equalization in the error-prone area according to the block constant, the detection image is obtained, which can improve the contrast of the single-channel pixel image and focus on reflecting the smaller local features inside the error-prone area, thereby improving the detection effect and reducing the complexity of calculation; It should be noted that when performing adaptive histogram equalization in normal areas, larger blocks are used, and when performing enhanced histogram equalization in error-prone areas, smaller blocks are used, which can reduce the amount of calculation in normal areas and improve the detection effect on error-prone areas.

[0023] The inspection image is compared with the single-channel historical pixel image, and the defect type is determined based on the similarity to classify the defects. When there is a defect, the inspection image is saved in the database to update the single-channel historical pixel image.

[0024] It should be noted that when making similarity judgments, methods such as local binary patterns and oriented gradient histograms may be used, which are all existing technologies and will not be described in detail here.

[0025] like Figure 2 As shown, the regional prediction method includes: Convert the historical defect map into a single-channel historical pixel map and mark the defect locations; Reassign the channel value of the defect position and change the channel value of the remaining positions to 0, obtain the initial area map, and assign weights to the initial area map according to the generation time of the single-channel historical pixel map; According to the weights of the initial area maps, the initial area maps are added in a matrix manner to obtain the predicted area map, specifically: ,in To predict the channel value of the area graph at the coordinate (x, y), is the channel value of the i-th initial area graph at the coordinate (x, y), is the weight of the i-th initial region map, which is set by the technicians. Its value is related to the generation time of its corresponding single-channel historical pixel map. The earlier the time, the smaller the value. is the total number of single-channel pixel images; In the forecast area diagram, ≥ The pixels are marked, and the area composed of all marked pixels is the error-prone area. The area division threshold is set by the technical staff and represents the size of the divided area.

[0026] It should be noted that, for ease of understanding, the simulation data after setting the assignment are shown in Table 1 and Table 2 below (the positions in the table represent the pixel positions, and the numbers represent the channel values ​​of the pixels). Table 1 and Table 2 represent different initial area maps respectively: Table 1: Initial Regions Figure 1

[0027] Table 2: Initial Regions Figure 2

[0028] To facilitate calculation, set all All are 1, is 0.5.

[0029] Calculate the value of each pixel based on the above data , as shown in Table 3 below: Table 3: Prediction areas Figure 1

[0030] The and By comparison, the area in the figure where the number is greater than 0.5 is the error-prone area, otherwise it is the normal area. The division of the error-prone area and the normal area can be achieved by matching the pixel position to the single-channel detection pixel map.

[0031] like Figure 5 As shown, adaptive histogram equalization includes: Divide the pixels in the normal area of ​​the single-channel detection pixel map into several non-overlapping blocks; Calculate each block separately: , ,in is the number of pixels with channel value i in the block, For the cumulative function, that is, the number of pixels in the block with channel values ​​i and below, The upper limit of the channel value in the single-channel detection pixel map, is the mapping channel value of pixel i in the single-channel detection pixel map in the detection map, is the upper limit of the channel value of the single-channel detection pixel map, and are the minimum and maximum values ​​in the cumulative function respectively; The detection map corresponding to the normal area can be obtained by mapping all the channel values ​​of the normal area in the single-channel detection pixel map.

[0032] It should be noted that, in order to facilitate calculation, The value is 3, and the normal area in the single-channel detection pixel map is shown in Table 4 below (the position in the table represents the pixel position, and the number represents the channel value of the pixel): Table 4: Normal area pixels Figure 1

[0033] The pixels in the normal area of ​​the above single-channel detection pixel map are divided into two non-overlapping blocks of size 2x3. The upper block is calculated according to the formula (the values ​​of i are 0, 1, 2, and 3 respectively): They are 2, 1, 3, and 0 respectively; They are 2, 3, 6, and 6 respectively; They are 0, 1, 3, and 3 (rounded off). Since the two blocks in the simulation data are exactly the same, the values ​​of the other blocks are consistent with the above calculation blocks. The normal area in the detection image is shown in Table 5 below: Table 5: Normal area pixels Figure 2

[0034] Compared with single-channel detection pixel images, the contrast is improved, making it easier to identify defect features. In actual use, the number of pixels is greater, so larger blocks can be used for calculations, reducing the amount of calculations.

[0035] like Figure 6 As shown, enhanced histogram equalization includes: Divide the pixels in the error-prone area in the single-channel detection pixel map into several non-overlapping blocks; Calculate each block separately: ,

[0036] in is the number of pixels with channel value i in the block, is a cumulative function, that is, the number of pixels with channel values ​​i and below in the block, The upper limit of the channel value in the single-channel detection pixel map, is the mapping channel value of pixel i in the single-channel detection pixel map in the detection map, is the upper limit of the channel value of the single-channel detection pixel map, and are the minimum and maximum values ​​in the cumulative function, respectively. is the block constant; The detection map corresponding to the error-prone area can be obtained by mapping all the channel values ​​of the error-prone area in the single-channel detection pixel map.

[0037] It should be noted that, in order to facilitate calculation, is 3, The error-prone areas in the single-channel detection pixel map are shown in Table 6 below (the positions in the table represent the pixel positions, and the numbers represent the channel values ​​of the pixels): Table 6: Pixels in error-prone areas Figure 1

[0038] The pixels in the error-prone area in the above single-channel detection pixel map are divided into four non-overlapping blocks of 2x2 size. The upper left corner block is calculated according to the formula (the values ​​of i are 0, 1, 2, and 3 respectively): They are 1, 1, 2, and 0 respectively; They are 1, 2, 4, 4 respectively; They are 0, 2, 3, and 3 respectively (the calculation result exceeds the upper limit of the channel value, and the maximum channel value is taken); since the four blocks in the simulation data are exactly the same, the values ​​of the remaining blocks are consistent with the above calculation blocks, and the error-prone areas in the detection map are obtained as shown in Table 7 below: Table 7: Pixels in error-prone areas Figure 2

[0039] Compared with the single-channel detection pixel map, the contrast is improved, and the defect features are easier to identify. In addition, compared with the normal area of ​​the detection image, a smaller block is used for calculation, and the channel value is further expanded by the block constant, which can reflect smaller local features and improve the accuracy of detection.

[0040] The block constant calculation method includes: Record the number of pixels whose calculated values ​​exceed the maximum number of channels in the last enhanced histogram equalization calculation; Get the total number of pixels in the error-prone area detected last time, and calculate the block constant, specifically: ,in is the block constant, is the block constant calculated last time, It is the original channel value whose channel value is greater than the maximum channel value after enhanced histogram equalization calculation in the previous calculation. is the upper limit of the channel value of the single-channel detection pixel map, is the number of pixels with channel value i in the error-prone area, Calculate the coefficient for the block, set by the technician, is the total number of pixels in the error-prone area detected last time.

[0041] It should be noted that in order to facilitate the calculation of the data using the enhanced histogram equalization, is 2, which means that in the last calculation, the corresponding channel values ​​of the pixels with channel values ​​of 2 and above in the single-channel detection pixel map exceed 3 after calculation. The number of pixels with channel values ​​of 2 in the error-prone area is 8, and the number of pixels with channel values ​​of 3 is 0. The total number of pixels in the error-prone area is 16. is 2, set is 1, so according to the formula , it can be calculated is 1.5.

[0042] It should be noted that in the enhanced histogram equalization calculation, when the calculated channel value exceeds the maximum channel value, the maximum channel value is taken. From the above data, it can be seen that when a part of the calculated channel values ​​exceeds the maximum channel value, the block constant of the next calculation can be reduced, thereby reducing the expansion of the channel value, preventing excessive assimilation of the channel value, and improving the detection accuracy.

[0043] Embodiment 2: In the first embodiment, the regional prediction method has less computational complexity, but is completely dependent on the previous position of the error-prone area, and can only divide the error-prone area according to the position of the defect that has already occurred. When the position offset of the defect is large, the defect position is easily divided outside the error-prone area, and the adaptability is poor. The difference between this embodiment and the first embodiment is that another regional prediction method is provided to reduce the dependence on the position of the previous error-prone area and further improve the accuracy of regional division.

[0044] Regional forecasting methods include: Convert the historical defect map into a single-channel historical pixel map, mark the defect location, and obtain the center pixel coordinates of the marked area; According to the central pixel coordinates of multiple defect positions, the coordinate relationship of the defect positions is established, specifically: ,in is the correlation coefficient between the horizontal and vertical coordinates of the center pixel coordinates, that is, the correlation coefficient between the x-coordinate and the y-coordinate. is the average value of the horizontal coordinates of each central pixel coordinate, is the horizontal coordinate value of the jth defect in the i-th single-channel historical pixel image, is the number of middle pixel coordinates of the i-th single-channel historical pixel image, is the ordinate value of the jth defect in the i-th single-channel historical pixel image, is the average value of the ordinate in each central pixel coordinate, is the total number of single-channel pixel images;

[0045] According to the correlation coefficient between the horizontal and vertical coordinates of the central pixel coordinates, a distribution prediction model is established to calculate the defect constant of each pixel position in the image to be tested, specifically: ,in is the defect constant of the pixel with coordinates (x, y) in the image to be detected, is the standard deviation of the horizontal coordinates of each central pixel coordinate, is the standard deviation of the ordinates in each central pixel coordinate, is the correlation coefficient between the horizontal and vertical coordinates of the center pixel coordinates, is the average value of the horizontal coordinates of each central pixel coordinate, is the average value of the ordinate in each central pixel coordinate, It is the abbreviation of exponential function, which is used to represent the exponential function with the natural constant e as the base. ; The distribution prediction model is used to calculate the coordinates of each pixel in the image to be detected. ≥ The pixels are marked, and the area composed of all marked pixels is the error-prone area. It is the set area division threshold, which represents the area size of the divided area.

[0046] It should be noted that, for the convenience of calculation, the simulated data are shown in Table 8 below: Table 8: Defect location table

[0047] calculate About 141, It is about 101. About 47, About 28, It is about -0.11. Select any coordinate in the single-channel detection pixel map to calculate the defect constant and compare it with For comparison, when the defect constant at the pixel position is greater than When , the pixel position can be marked as an error-prone area, otherwise it can be marked as a normal area. After calculating all pixel positions, the normal area and the error-prone area can be divided. For example, (160,110) and (100,160), f(100,160) is approximately 1.04x10 -5 , the calculated value of f(160, 110) is approximately 1.036x10 -4 ,set up 1x10 -4 , then f(100, 160) is less than , indicating that the probability of a defect is small, marking the pixel position as a normal area, and f (160, 110) is greater than , indicating that the probability of a defect is high, marking the pixel position as an error-prone area.

[0048] Compared with the first embodiment, the calculation and division results of the second embodiment are more accurate, and can allow for greater fluctuations in the defect position, so the calculation results are more accurate.

[0049] Embodiment three: In the second embodiment, the region is divided according to the center coordinates of the defect, and the region division threshold It represents the divided area of ​​the defect. However, in actual production, the area of ​​the defect may change frequently. Therefore, the division method in Example 2 may cause the error-prone area to be too large or too small, which is not conducive to detection. Based on Example 2, this embodiment provides a threshold changing method to further improve the accuracy of the error-prone area division.

[0050] The threshold iteration methods include: Check whether the defect position of the image to be inspected exceeds the error-prone area; When the defect position exceeds the error-prone area, detect whether the defect position exceeds the error-prone area in the positive and negative directions of the horizontal axis or the positive and negative directions of the vertical axis at the same time. When it exceeds the error-prone area at the same time, obtain the maximum distance of the excess and calculate the change in the area division threshold, specifically: ,in The threshold for setting the area division. The area division threshold calculated last time, is the maximum distance beyond the error prone zone, It is the update coefficient, which is set by the technicians and its value represents the sensitivity of the update amplitude; When the defect position does not exceed the error-prone area, the minimum distance between the edge of the defect position and the edge of the error-prone area is obtained, and the change in area division is calculated, specifically: ,in Indicates the number of times the area division threshold has been increased. The threshold for setting the area division. The area division threshold calculated last time, is the minimum distance between the edge of the defect position and the edge of the error-prone area, It is the update coefficient, which is set by the technician and its value represents the sensitivity of the update amplitude.

[0051] It should be noted that in order to facilitate calculation, the simulation data is set as follows: The defect position does not exceed the error-prone area, and the minimum distance between the edge of the defect position and the edge of the error-prone area is 2000 pixels. is 2, 1x10 -4 , Using Example 2 The data is 1x10 -4 ; Calculation can be obtained 1.2x10 -4 .

[0052] It can be seen from this that when the defect area is small, the regional division threshold can be increased, thereby reducing the number of defect constants greater than the regional division threshold, thereby narrowing the scope of the error-prone area, further improving the accuracy of the division of the error-prone area, reducing the amount of calculation, and improving the detection efficiency. Similarly, when the defect area is large, the regional division threshold can be lowered, the scope of the error-prone area can be expanded, and the detection accuracy can be improved.

[0053] Embodiment 4: In Example 1, when calculating the error-prone area through enhanced histogram equalization, the blocks are calculated separately, which may cause the blocks to be not smooth enough and affect the detection effect. This embodiment provides an optimization method to further optimize the mapping channel value of the block edge and improve the smoothness between blocks.

[0054] Optimization methods include: Mark all block edge pixels, calculate the mapping channel value of the target pixel in the adjacent block, and perform weighted summation of the mapping channel value according to the number of adjacent pixels in the corresponding adjacent block, specifically: ,in is the mapping channel value of the adjacent block z (including the block where the target pixel is located) to the target pixel, Optimize values ​​for edge pixels, is the number of adjacent pixels of the target pixel in the adjacent block z (including the block where the target pixel is located), is the total number of adjacent pixels of the target pixel; After calculating the edge pixel optimization values ​​of all block edge pixels, the channel values ​​of the target pixels are reassigned using the edge pixel optimization values ​​to obtain the optimized detection image.

[0055] It should be noted that, for the convenience of calculation, the data of enhanced histogram equalization in Example 1 is used: Take the lower right pixel of the upper left corner block in Table 7 as an example. Its channel value in Table 6 is 2. It has two adjacent pixels in the upper left corner block ( is 2), and the mapping channel value of the upper left corner block for the channel value 2 is 3 ( is 3); it has one neighboring pixel in the upper right corner block ( is 1), and the mapping channel value of the pixel with channel value 2 in the upper right corner is 3 ( is 3), which has a neighboring pixel in the lower left corner block ( is 1), and the mapping channel value of the pixel with channel value 2 in the lower left corner block is 3 ( =3), according to the formula , it can be calculated that the edge pixel optimization value of the pixel The mapping channel value at the edge of the block is recalculated to improve the smoothness between blocks, thereby improving the detection accuracy.

[0056] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is limited by the attached embodiments and their equivalents.

Claims

1. A feedback-driven chip defect classification method based on visual analysis, characterized by: include: Establish a database, collect historical defect images and record the time when the historical defect images were generated, convert the historical defect images into single-channel historical pixel images, and classify them according to the different labels of the defects; Obtain the image to be detected, convert the image to be detected into a single-channel detection pixel map, divide the single-channel detection pixel map into normal areas and error-prone areas by using the regional prediction method according to the historical defect images, and calculate the block constant of the error-prone area by using the block constant calculation method; Adopting blocks of different sizes, performing adaptive histogram equalization in the normal area, and performing enhanced histogram equalization in the error-prone area according to the block constant, the detection image is obtained, which can improve the contrast of the single-channel pixel image and focus on reflecting the smaller local features inside the error-prone area, thereby improving the detection effect and reducing the complexity of calculation; The inspection image is compared with the single-channel historical pixel image, and the defect type is determined based on the similarity to classify the defects. When there is a defect, the inspection image is saved in the database to update the single-channel historical pixel image.

2. The feedback driver chip defect classification method based on visual analysis according to claim 1 is characterized in that: The regional prediction method includes: Convert the historical defect map into a single-channel historical pixel map and mark the defect locations; Reassign the channel value of the defect position and change the channel value of the remaining positions to 0, obtain the initial area map, and assign weights to the initial area map according to the generation time of the single-channel historical pixel map; According to the weights of the initial area maps, the initial area maps are added in a matrix manner to obtain the predicted area map, specifically: ,in To predict the channel value of the area graph at the coordinate (x, y), is the channel value of the i-th initial area graph at the coordinate (x, y), is the weight of the i-th initial region map, which is set by the technicians. Its value is related to the generation time of its corresponding single-channel historical pixel map. The earlier the time, the smaller the value. is the total number of single-channel pixel images; In the forecast area diagram, ≥ The pixels are marked, and the area composed of all marked pixels is the error-prone area. The area division threshold is set by the technical staff and represents the size of the divided area.

3. The feedback driver chip defect classification method based on visual analysis according to claim 1 is characterized in that: The regional prediction method includes: Convert the historical defect map into a single-channel historical pixel map, mark the defect location, and obtain the center pixel coordinates of the marked area; According to the central pixel coordinates of multiple defect positions, the coordinate relationship of the defect positions is established, specifically: ,in is the correlation coefficient between the horizontal and vertical coordinates of the center pixel coordinates, that is, the correlation coefficient between the x-coordinate and the y-coordinate. is the average value of the horizontal coordinates of each central pixel coordinate, is the horizontal coordinate value of the jth defect in the i-th single-channel historical pixel image, is the number of middle pixel coordinates of the i-th single-channel historical pixel image, is the ordinate value of the jth defect in the i-th single-channel historical pixel image, is the average value of the ordinate in each central pixel coordinate, is the total number of single-channel pixel images; According to the correlation coefficient between the horizontal and vertical coordinates of the central pixel coordinates, a distribution prediction model is established to calculate the defect constant of each pixel position in the image to be tested, specifically: ,in is the defect constant of the pixel with coordinates (x, y) in the image to be detected, is the standard deviation of the horizontal coordinates of each central pixel coordinate, is the standard deviation of the ordinates in each central pixel coordinate, is the correlation coefficient between the horizontal and vertical coordinates of the center pixel coordinates, is the average value of the horizontal coordinates of each central pixel coordinate, is the average value of the ordinate in each central pixel coordinate, It is the abbreviation of exponential function, which is used to represent the exponential function with the natural constant e as the base. ; The distribution prediction model is used to calculate the coordinates of each pixel in the image to be detected. ≥ The pixels are marked, and the area composed of all marked pixels is the error-prone area. It is the set area division threshold, which represents the area size of the divided area.

4. The feedback driver chip defect classification method based on visual analysis according to claim 3 is characterized in that: The area division threshold can be updated according to a threshold iteration method, and the threshold iteration method includes: Check whether the defect position of the image to be inspected exceeds the error-prone area; When the defect position exceeds the error-prone area, detect whether the defect position exceeds the error-prone area in the positive and negative directions of the horizontal axis or the positive and negative directions of the vertical axis at the same time. When it exceeds the error-prone area at the same time, obtain the maximum distance of the excess and calculate the change in the area division threshold, specifically: ,in The threshold for setting the area division. The area division threshold calculated last time, is the maximum distance beyond the error prone zone, It is the update coefficient, which is set by the technicians and its value represents the sensitivity of the update amplitude; When the defect position does not exceed the error-prone area, the minimum distance between the edge of the defect position and the edge of the error-prone area is obtained, and the change in area division is calculated, specifically: ,in Indicates the number of times the area division threshold has been increased. The threshold for setting the area division. The area division threshold calculated last time, is the minimum distance between the edge of the defect position and the edge of the error-prone area, It is the update coefficient, which is set by the technician and its value represents the sensitivity of the update amplitude.

5. The feedback driver chip defect classification method based on visual analysis according to claim 1 is characterized in that: The block constant calculation method includes: Record the number of pixels whose calculated values ​​exceed the maximum number of channels in the last enhanced histogram equalization calculation; Get the total number of pixels in the error-prone area detected last time, and calculate the block constant, specifically: ,in is the block constant, is the block constant calculated last time, It is the original channel value whose channel value is greater than the maximum channel value after enhanced histogram equalization calculation in the previous calculation. is the upper limit of the channel value of the single-channel detection pixel map, is the number of pixels with channel value i in the error-prone area, Calculate the coefficient for the block, set by the technician, is the total number of pixels in the error-prone area detected last time.

6. The feedback driver chip defect classification method based on visual analysis according to claim 1 is characterized in that: The adaptive histogram equalization comprises: Divide the pixels in the normal area of ​​the single-channel detection pixel map into several non-overlapping blocks; Calculate each block separately: , ,in is the number of pixels with channel value i in the block, For the cumulative function, that is, the number of pixels in the block with channel values ​​i and below, The upper limit of the channel value in the single-channel detection pixel map, is the mapping channel value of pixel i in the single-channel detection pixel map in the detection map, is the upper limit of the channel value of the single-channel detection pixel map, and are the minimum and maximum values ​​in the cumulative function respectively; The detection map corresponding to the normal area can be obtained by mapping all the channel values ​​of the normal area in the single-channel detection pixel map.

7. The feedback driver chip defect classification method based on visual analysis according to claim 1 is characterized in that: The enhanced histogram equalization comprises: Divide the pixels in the error-prone area in the single-channel detection pixel map into several non-overlapping blocks; Calculate each block separately: , ,in is the number of pixels with channel value i in the block, is a cumulative function, that is, the number of pixels with channel values ​​i and below in the block, The upper limit of the channel value in the single-channel detection pixel map, is the mapping channel value of pixel i in the single-channel detection pixel map in the detection map, is the upper limit of the channel value of the single-channel detection pixel map, and are the minimum and maximum values ​​in the cumulative function, respectively. is the block constant; The detection map corresponding to the error-prone area can be obtained by mapping all the channel values ​​of the error-prone area in the single-channel detection pixel map.

8. The feedback driver chip defect classification method based on visual analysis according to claim 7 is characterized in that: The enhanced histogram equalization also includes an optimization method for block edges, specifically: Mark all block edge pixels, calculate the mapping channel value of the target pixel in the adjacent block, and perform weighted summation of the mapping channel value according to the number of adjacent pixels in the corresponding adjacent block, specifically: ,in is the mapping channel value of the adjacent block z to the target pixel, Optimize values ​​for edge pixels, is the number of neighboring pixels of the target pixel in the adjacent block z, is the total number of adjacent pixels of the target pixel; After calculating the edge pixel optimization values ​​of all block edge pixels, the channel values ​​of the target pixels are reassigned using the edge pixel optimization values ​​to obtain the optimized detection image.

9. Feedback-driven chip defect classification system based on visual analysis, characterized by: include: Data storage module: used to establish a database, collect historical defect images and record the generation time of historical defect images, convert historical defect images into single-channel historical pixel images, and classify labels according to different defects; Data sorting module: obtain the image to be tested, convert it into a single-channel detection pixel map, divide the single-channel detection pixel map into normal areas and error-prone areas through the regional prediction method according to the historical defect images, and calculate the block constant of the error-prone area through the block constant calculation method; Data processing module: using blocks of different sizes, performing adaptive histogram equalization in the normal area, and performing enhanced histogram equalization in the error-prone area according to the block constant to obtain the detection image, which can improve the contrast of the single-channel pixel image and focus on reflecting the smaller local features inside the error-prone area, thereby improving the detection effect and reducing the complexity of the calculation; Data output module: compares the inspection image with the single-channel historical pixel image, determines the defect type based on the similarity, and classifies the defects. When there is a defect, saves the inspection image to the database and updates the single-channel historical pixel image.

Images