Computer vision-based integrated circuit board chip surface defect recognition method

Abstract

The application provides a computer vision-based integrated circuit board chip surface defect recognition method, and belongs to the field of image processing. The method comprises the following steps: acquiring a chip image, clustering pixel points to obtain clustering clusters, and then obtaining regions and edges in each component; dividing the edges of each component into a plurality of local edges; comparing different local edges of each component to obtain shape defect regions and shape normal regions; dividing the shape normal regions in each component into a plurality of equal regions to obtain to-be-analyzed regions in each component; comparing the distribution of pixel point gray values in different to-be-analyzed regions in the same component to obtain regions containing surface defects. The application aims to solve the problem that current defect detection methods cannot be well applied to all components in a chip due to the great difference between the shapes and colors of different components in the chip.

Description

Technical Field

[0001] This invention belongs to the field of image processing, specifically relating to a method for identifying surface defects of integrated circuit board chips based on computer vision. Background Technology

[0002] As electronic devices become increasingly complex and integrated, defect detection of circuit board chips is becoming increasingly critical. Defects directly affect the function and performance of circuit boards. Timely detection and repair of these defects can significantly improve the yield rate of circuit boards, reduce repair costs and time caused by failures, and thus improve production efficiency and economic benefits. With the development of image processing technologies, the application of image processing technology for defect identification is becoming increasingly widespread. Image processing technology can quickly and accurately analyze image data on the surface of circuit board chips to identify appearance defects. Therefore, computer vision is used to identify surface defects of integrated circuit board chips.

[0003] Because different components within a chip have significant differences in color and shape, current methods for detecting shape and surface defects on the chip surface, such as threshold segmentation or other methods, are not applicable to all critical components within the chip. This results in significant discrepancies between the detected shape and surface defect information and the actual situation. Summary of the Invention

[0004] To address the issue that current detection methods are not well-suited for all critical components within a chip due to significant differences in color and shape among different internal components, this invention proposes a computer vision-based method for identifying surface defects on integrated circuit board chips.

[0005] To achieve the above objectives, the present invention provides the following technical solution: acquiring an image of a chip package; clustering the pixels in the chip package image according to the grayscale values ​​of the pixels to obtain several clusters; recording the pixels in each cluster as pixels in each component; obtaining the regions and edges of each component based on the pixels in each component; acquiring the corner points on the edges of each component, and dividing the edges of each component into several local edges based on the corner points on the edges of each component; obtaining the length and tilt angle of the local edges of each component based on the distribution and number of pixels on the local edges of each component; obtaining the degree of defect manifestation of the local edges of each component based on the difference in the length and tilt angle of different local edges of the same component, and further dividing the regions in all components into shape defect regions and normal shape regions; dividing the normal shape regions in each component into multiple equally sized regions to obtain the region to be analyzed in each component; obtaining the degree of surface defects in each region to be analyzed in each component based on the difference in information entropy of the grayscale values ​​of pixels in different regions to be analyzed in the same component and the difference in the grayscale values ​​of pixels at the same position in different regions to be analyzed in the same component, and further obtaining the region containing surface defects, thus completing defect detection.

[0006] Furthermore, the specific steps for obtaining several clusters are as follows: using the elbow method, the optimal number of clusters is obtained when clustering all pixels in the chip package image using the K-means clustering algorithm. Let the number of clusters be Based on the grayscale value of each pixel in the chip package image, the K-means clustering algorithm is used to cluster all pixels in the chip package image to obtain... There are several clusters.

[0007] Furthermore, the specific steps for obtaining the region within each component and the edge of each component are as follows: According to the first... The pixels within a component are used to obtain several regions, which are denoted as the nth region. The area within each component; the first The first component The edge of the i-th region is denoted as the i-th region. The edge of each component.

[0008] Furthermore, the specific steps for dividing the edge of each component into several local edges of each component are as follows: through Corner detection algorithms acquire all corner points within the chip package image, obtaining the first... The corner points on the edge of the component; the first The corner points on the edges of each component are used as segmentation points to divide the first component into segments. Each edge of a component is divided into multiple segments, and each segment is denoted as the first segment. Local edges of individual components.

[0009] Furthermore, the specific steps for obtaining the length and tilt angle of the local edge of each component are as follows: Obtain the first... The first component The 8-adjacent chain code of the local edge; if the first The first component If the most frequent number in the 8-adjacent chain code of a local edge is less than 4, then the 1st... The first component The angle corresponding to the most frequent number in the 8-adjacent chain codes of a local edge is denoted as the i-th. The first component The tilt angle of the local edge; if the first The first component If the number that appears most frequently in the 8-adjacent chain codes of a local edge is greater than or equal to 4, then the 1st... The first component The angle corresponding to the value obtained by subtracting 4 from the most frequent number in the 8-adjacency chain code of a local edge is denoted as . The first component The tilt angle of the local edge; the first The first component The number of pixels contained on a local edge is denoted as the i-th. The first component The length of a local edge.

[0010] Furthermore, the specific calculation formula for obtaining the degree of defect manifestation at the local edge of each component is as follows: ; In the formula, Indicates the first The first component The degree of defect manifestation at a local edge, Indicates the first The first component The local edge and the first Differences in local edges, Indicates the first The first component The length of a local edge, Indicates the first The first component The length of a local edge, Indicates the first The maximum value among the lengths of all local edges of a component. Indicates the first The first component The tilt angle of a local edge. Indicates the first The first component The tilt angle of a local edge. Indicates the first The maximum value of the tilt angles of all local edges of a component. Indicates the first The number of local edges of each component Represents the absolute value function. express function, express function, It is an exponential function with the natural constant as its base.

[0011] Furthermore, the specific steps for dividing the areas within all components into shape defect areas and normal shape areas are as follows: Preset a threshold for the degree of defect manifestation. ;like Then the first The first component The region containing a local edge is denoted as the shape defect region; if Then the first The first component The region containing a local edge is denoted as the normally shaped region. Indicates the first The first component The degree of defect manifestation at a local edge.

[0012] Furthermore, the specific steps for obtaining the region to be analyzed within each component are as follows: If the first... If the number of normally shaped areas within a component is 1, then... The normally shaped area within a component is divided into several identically sized... A rectangular region, denoted as the first... The region to be analyzed within each component; among which... The width of the preset rectangular area; if the first If the number of normally shaped regions within a component is greater than 1, then the first component will be... The normally shaped area within a component is denoted as the [number]th [component]. The area to be analyzed within each component.

[0013] Furthermore, the specific calculation formula for obtaining the surface defect degree of each region to be analyzed within each component is as follows: In the formula, Indicates the first Within the first component The degree of surface defects in the area to be analyzed Indicates the first Within the first component The number of pixels contained in the region to be analyzed. Indicates the first Within the first component The first region to be analyzed The grayscale value of each pixel Indicates the first Within each component, in all regions to be analyzed, the first... The average grayscale value of each pixel Indicates the first The number of regions to be analyzed within each component. Indicates the first Within the first component The information entropy of the grayscale values ​​of pixels in the region to be analyzed. Indicates the first Within the first component The information entropy of the grayscale values ​​of pixels in the region to be analyzed. Indicates the first Within the first component The region to be analyzed and the first The absolute value of the difference in the number of pixels in the region to be analyzed. To prevent hyperparameters with a denominator of 0, Indicates the first The variance of the information entropy of the grayscale values ​​of pixels in all regions to be analyzed within a component. Represents the absolute value function. express function, express function.

[0014] Furthermore, the specific steps for obtaining the region containing surface defects are as follows: Preset a surface defect severity threshold. ,like Then the first Within the first component The region to be analyzed is denoted as the region containing surface defects. Indicates the first Within the first component The degree of surface defects in the area to be analyzed.

[0015] The computer vision-based integrated circuit board chip surface defect identification method provided by this invention has the following beneficial effects: When performing defect detection on a chip, this invention first divides the chip package into multiple components based on the characteristic that the gray values ​​of pixels within the same component are relatively similar. Based on the certain rules followed during chip manufacturing, different local edges within the same component normally exhibit strong similarity. By comparing multiple local edges within a component, areas containing shape defects are obtained. This solves the problem that current detection methods cannot effectively detect shape defects in all components within a chip due to differences in shape and color between different components within the chip package. Furthermore, based on the characteristic that chip components are normally solid colors, the distribution of gray values ​​of pixels on the same component at different locations within the chip is relatively similar. By comparing different locations within each component or comparing the same components at different locations, areas with surface defects are obtained. This solves the problem that current detection methods cannot effectively detect surface defects in all components within a chip due to significant differences in shape or color between different components within the chip. Attached Figure Description

[0016] To more clearly illustrate the embodiments and design schemes of the present invention, the accompanying drawings required for this embodiment will be briefly described below. The drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of a computer vision-based method for identifying surface defects on integrated circuit board chips, according to an embodiment of the present invention. Detailed Implementation

[0018] To enable those skilled in the art to better understand and implement the technical solutions of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be construed as limiting the scope of protection of the present invention.

[0019] Example 1: This invention provides a method for identifying surface defects of integrated circuit board chips based on computer vision, specifically as follows: Figure 1 As shown, it includes: Step S001: Obtaining an image of the chip package.

[0020] Specifically, while ensuring the camera's optical axis is perpendicular to the chip package surface, a high-resolution industrial camera is used to photograph the chip package surface, obtaining an initial image of the chip package. A ring light, with white light color, is used as the light source when photographing the chip package.

[0021] Furthermore, the initial image of the chip package is converted to grayscale, and the resulting image is recorded as the chip package image. The grayscale conversion operation is a well-known existing technique and will not be described in detail in this embodiment.

[0022] At this point, the chip package image is obtained.

[0023] Step S002: Based on the grayscale values ​​of the pixels, cluster the pixels in the chip package image to obtain several clusters; record the pixels in each cluster as pixels in each component; based on the pixels in each component, obtain the region and edge of each component; obtain the corner points on the edge of each component, and divide the edge of each component into several local edges based on the corner points on the edge of each component; based on the distribution and number of pixels on the local edges of each component, obtain the length and tilt angle of the local edges of each component; based on the difference in the length and tilt angle of different local edges of the same component, obtain the degree of defect manifestation of the local edges of each component, and then divide the region in all components into shape defect region and normal shape region.

[0024] It's important to note that under normal circumstances, during chip manufacturing, the same production standards are applied to the same parts, resulting in strong similarity within the same part of a chip. For example, each normally exposed pin inside a chip is highly similar to other normally exposed pins. However, when a chip has a defect, the defective part will differ from the normal part. Therefore, the chip package is disassembled. Then, by comparing each part within the disassembled chip package with other parts, defect detection is performed on the chip package.

[0025] It's important to further explain that a chip package mainly consists of multiple components such as the package shell, lead frame, and leads. Furthermore, components of the same type on the same chip package typically have the same style. In a chip package image, the most direct representation of style includes shape and color. This means that even if some components have shape defects, the colors of identical components will be the same; that is, the grayscale values ​​of pixels on the same component are quite similar in a grayscale image. Therefore, a clustering algorithm is used to cluster pixels based on their grayscale values, dividing the pixels into multiple clusters to complete the decomposition of the chip package.

[0026] It should be further explained that, since the same normal component in different locations within a chip package is quite similar—that is, the same edges of the same component in different locations within each chip package are quite similar—it is necessary to obtain the edge of each component in the chip package and then compare the similarity of the edges within a cluster to determine the degree of defect for each edge.

[0027] It should be further explained that since some components within a chip consist of only one unit, such as the package shell (i.e., the chip shell), it is impossible to directly compare the edges of the package shell with the edges of other components within the same chip to detect shape defects in the chip package. Furthermore, because the qualified components within a chip exhibit strong regularity—meaning each edge segment of a qualified component shows strong similarity to other edge segments—the edge of each component is divided into multiple local edges. The degree of defect for each local edge is determined based on the difference in tilt angle and length between each local edge and other local edges.

[0028] It should be further explained that when obtaining the tilt angle of each local edge, since some edges are not straight segments (i.e., some edges have no tilt angle), the tilt angle of each edge is obtained through the edge chain code, and this tilt angle is used to replace the tilt angle of each edge. Since the tilt angle difference of 4 values ​​in the 8 critical chain codes corresponds to a difference of 180 degrees, meaning the straight lines obtained from the tilt angles corresponding to two chain code values ​​differing by 4 are parallel, chain code values ​​greater than 4 need to be modified when obtaining the tilt angle through the edge chain code.

[0029] It should be further explained that the point where two lines intersect is also called a corner point. Therefore, by using a corner point detection algorithm to obtain the corner points within the chip package image, the corner points on the edge are used as segmentation points to divide the edge into multiple local edges.

[0030] It should be further noted that some local edges within a chip package image are not necessarily similar to all local edges. For example, even within a rectangular region, the long and short sides can differ significantly. Therefore, when comparing different local edges within a component, different weights are assigned to the comparison results of different local edges.

[0031] Specifically, the elbow method is used to obtain the optimal number of clusters when clustering all pixels within the chip package image using the K-means clustering algorithm. Based on the grayscale value of each pixel in the chip package image, the K-means clustering algorithm is used to cluster all pixels in the chip package image, resulting in... There are several clusters. The methods of obtaining the optimal number of clusters using the elbow method and clustering the data using the K-means clustering algorithm are both well-known existing techniques and will not be described in detail in this embodiment.

[0032] Furthermore, the first The pixels within the i-th cluster are denoted as the i-th... The pixels of each component. According to the first... The pixels within a component are used to obtain several regions, which are denoted as the nth region. The area within each component. The first... The first component The edge of the i-th region is denoted as the i-th region. The edge of each component.

[0033] At this point, the edges of each component are obtained.

[0034] Furthermore, through A corner detection algorithm is used to obtain all corner points within the chip package image. The first... The corner points on the edges of each component are used as segmentation points to divide the first component into segments. Each edge of a component is divided into multiple segments, and each segment is denoted as the first segment. The local edges of each component. The method of obtaining corner points within the image using a corner detection algorithm is a well-known existing technique and will not be described in detail in this embodiment.

[0035] Furthermore, obtain the first The first component The 8-adjacency chain code of a local edge. If the first... The first component If the most frequent number in the 8-adjacent chain code of a local edge is less than 4, then the 1st... The first component The angle corresponding to the most frequent number in the 8-adjacent chain codes of a local edge is denoted as the i-th. The first component The tilt angle of a local edge. If the first... The first component If the number that appears most frequently in the 8-adjacent chain codes of a local edge is greater than or equal to 4, then the 1st... The first component The angle corresponding to the value obtained by subtracting 4 from the most frequent number in the 8-adjacency chain code of a local edge is denoted as . The first component The tilt angle of a local edge. The first... The first component The number of pixels contained on a local edge is denoted as the i-th. The first component The length of a local edge. Obtaining the chain code for an edge is a well-known technique and will not be described in detail in this embodiment.

[0036] Furthermore, obtain the first The first component The specific formula for calculating the degree of defect manifestation at a local edge is as follows: ; In the formula, Indicates the first The first component The degree of defect manifestation at a local edge, Indicates the first The first component The local edge and the first Differences in local edges, Indicates the first The first component The length of a local edge, Indicates the first The first component The length of a local edge, Indicates the first The maximum value among the lengths of all local edges of a component. Indicates the first The first component The tilt angle of a local edge. Indicates the first The first component The tilt angle of a local edge. Indicates the first The maximum value of the tilt angles of all local edges of a component. Indicates the first The number of local edges of each component Represents the absolute value function. express The function whose input is the first... The first component Each local edge and every local edge ; express The function, used in this embodiment, is for normalization processing; This is an exponential function with the natural constant as its base, and in this embodiment it is used for inverse proportional processing.

[0037] It should be noted that, The smaller the value, the higher the value of the first element in the chip package image. The first component The local edge and the first The local edges are quite similar, the first... Each local edge conforms to the characteristic that each normal local edge within the chip package is relatively similar to other normal local edges, i.e., the first local edge... The first component The greater the probability that a local edge is a normal edge; The smaller the value, the more it indicates that the chip package image contains the first [value] under normal circumstances. The first component The local edge and the first The local edges are also quite likely to be similar, therefore, through get The weights, such that if A smaller value indicates that the two edges are more likely to be similar under normal circumstances; increasing the value... To mitigate the impact on the final result, this study aims to reduce the problem that the result has a significant influence on the final result when comparing normal edges that have certain differences at different positions within normal components.

[0038] Furthermore, if Then the first The first component The region containing a local edge is denoted as the shape defect region; if Then the first The first component The region containing a local edge is denoted as a region with normal shape. In this embodiment, a preset threshold for the degree of defect manifestation is used. This example is used to illustrate the concept; other values ​​can be set in other implementations.

[0039] Thus, the defective and normal shape regions within the chip package image are obtained.

[0040] Step S003: Divide the normal-shaped area within each component into multiple equally sized areas to obtain the area to be analyzed within each component; based on the difference in information entropy of the gray values ​​of pixels in different areas to be analyzed within the same component and the difference in gray values ​​of pixels at the same position in different areas to be analyzed within the same component, obtain the degree of surface defects in each area to be analyzed within each component, and thus obtain the area containing surface defects, completing the defect detection.

[0041] It should be noted that, due to the internal structure of the chip... Junctions can produce a photovoltaic effect under illumination, which can interfere with the normal operation of circuits. Black encapsulation materials, such as epoxy resin containing carbon powder, can effectively block external light sources, avoiding malfunctions caused by related negative effects. Furthermore, a uniform, solid-color surface facilitates even heat distribution, preventing localized overheating. This ensures that every component within the chip package is a solid color. That is, the grayscale distribution of pixels in normal components at different locations within the chip package is relatively similar. However, when components within the chip have scratches or corrosion areas, the pixel distribution at the scratched or corroded areas differs from the pixel distribution in normally shaped areas. Therefore, comparing different areas within each component reveals the probability of surface defects in each area.

[0042] It should be further explained that since some components within the chip package consist of only one area, such as the package shell, this component with only one area is divided into multiple local areas of the same size. By comparing the distribution of pixel grayscale values ​​in each local area with those in other local areas, the probability of surface defects in each local area can be determined.

[0043] It should be further explained that, since information entropy reflects the distribution of pixel grayscale values ​​within a region, the difference in information entropy between the pixel grayscale values ​​of each region and other regions within the same component is used as a parameter to calculate the probability of surface defects in each region. Because the number of pixels in each region differs from that in other regions, and because a larger number of pixels in a region may indicate a more disordered distribution of grayscale values, when calculating the information entropy of pixel grayscale values ​​within a region, the weights of the information entropy of pixel grayscale values ​​in the two regions are assigned based on the difference in the number of pixels in each region when comparing the information entropy of different regions to determine the probability of surface defects in each region.

[0044] It should be further noted that if all regions within the same component are normally shaped, then the information entropy of the pixel grayscale values ​​in all regions within that component will be relatively similar. In other words, the more stable the information entropy distribution of the comparison regions, the higher the similarity between the comparison regions, and the stronger the reliability of the comparison results between each region within that component and other regions. Therefore, the reciprocal of the variance of the information entropy of the pixel grayscale values ​​in all regions within each component is used as a premise for comparing the information entropy of each region with that of other regions.

[0045] It should be further explained that the difference in grayscale values ​​of pixels at the same location in two different regions reflects the difference in the distribution of grayscale values ​​of pixels in these two regions. Therefore, the difference in grayscale values ​​of pixels at the same location in different regions within the same component is used as a calculation parameter to determine the probability of surface defects in a local area.

[0046] Specifically, if the first If the number of normally shaped regions within a component is greater than 1, then the first component will be... The normally shaped area within a component is denoted as the [number]th [component]. The area to be analyzed within each component.

[0047] Furthermore, if the first If the number of normally shaped regions within a component is 1, then the first component will be... The normally shaped area within a component is divided into several identically sized... A rectangular region, denoted as the first... The analysis area is located within each component. When segmenting the cable surface area, if the number of pixels in a portion of the cable surface area is insufficient to fill the rectangular area, it is not padded. Furthermore, different rectangular areas do not contain the same number of pixels. The preset width of the rectangular area in this embodiment... This example is used to illustrate the concept; other values ​​can be set in other implementations.

[0048] Specifically, to obtain the first Within the first component The specific formula for calculating the degree of surface defects in the area to be analyzed is as follows: In the formula, Indicates the first Within the first component The degree of surface defects in the area to be analyzed Indicates the first Within the first component The number of pixels contained in the region to be analyzed. Indicates the first Within the first component The first region to be analyzed The grayscale value of each pixel Indicates the first Within each component, in all regions to be analyzed, the first... The average grayscale value of each pixel Indicates the first The number of regions to be analyzed within each component. Indicates the first Within the first component The information entropy of the grayscale values ​​of pixels in the region to be analyzed. Indicates the first Within the first component The information entropy of the grayscale values ​​of pixels in the region to be analyzed. Indicates the first Within the first component The region to be analyzed and the first The absolute value of the difference in the number of pixels in the region to be analyzed. Indicates the first The variance of the information entropy of the grayscale values ​​of pixels in all regions to be analyzed within a component. Represents the absolute value function. To prevent hyperparameters with a denominator of 0, this embodiment sets... Other values ​​may be set in other embodiments; express The function whose input is the first... Within the first component The region to be analyzed and the component excluding the first The reciprocal of the absolute value of the difference in the number of pixels in each region to be analyzed outside the region to be analyzed; express The function, used in this embodiment, is for normalization processing. The method of obtaining the information entropy of the grayscale values ​​of pixels within a region is a well-known existing technique and will not be elaborated upon in this embodiment.

[0049] It should be noted that, The larger the value, the more significant the first... Within the first component The greater the difference in grayscale values ​​between the region to be analyzed and the pixels at the same location in other regions within the same component, the more it indicates that the... Within the first component The distribution of pixel grayscale values ​​in the region to be analyzed differs significantly from that in other regions within the same component, i.e., the first region... Within the first component The greater the likelihood that the surface of the area to be analyzed contains defects; The larger the value, the more significant the [value]. Within the first component The region to be analyzed and the first The closer the number of pixels in each region to be analyzed, the more similar they are. The stronger the reference value, the better at this time. Assign a larger weight; The larger the value, the more significant the first... Within the first component The distribution of pixel grayscale values ​​in the region to be analyzed differs significantly from that in other regions of the same component, i.e., the first region... Within the first component The greater the likelihood that the surface of the area to be analyzed contains defects; The larger the value, the more significant the first... The more areas within a component that are normally shaped, the greater the likelihood that it will be. The stronger the credibility.

[0050] Furthermore, a preset threshold for the degree of surface defects is established. ,like Then the first Within the first component The region to be analyzed is denoted as the region containing surface defects. In this embodiment, a preset surface defect threshold is used. This example is used to illustrate the concept; other values ​​can be set in other implementations.

[0051] This reveals the region containing surface defects within the chip package image.

Claims

1. A method for identifying surface defects of integrated circuit board chips based on computer vision, characterized in that, include: Obtain an image of the chip package; Based on the grayscale value of the pixels, the pixels in the chip package image are clustered to obtain several clusters; Each pixel within a cluster is recorded as a pixel within a component. Based on the pixels within each component, the region and edge of each component are obtained. Corner points on the edges of each component are acquired, and based on these corner points, the edges of each component are divided into several local edges. The length and tilt angle of each local edge are obtained based on the distribution and number of pixels on its local edges. The degree of defect manifestation of each component's local edges is determined based on the differences in length and tilt angle among different local edges of the same component. This process further divides the regions within all components into shape-defect regions and shape-normal regions. The shape-normal regions within each component are then divided into multiple equally sized regions to obtain the region to be analyzed within each component. The degree of surface defect in each region to be analyzed within each component is obtained based on the differences in information entropy of pixel grayscale values ​​in different regions to be analyzed within the same component and the differences in grayscale values ​​of pixels at the same position in different regions to be analyzed within the same component. This process yields the region containing surface defects, completing defect detection. The specific calculation formula for obtaining the degree of defect manifestation of each component's local edges is as follows: ; In the formula, Indicates the first The first component The degree of defect manifestation at a local edge, Indicates the first The first component The local edge and the first Differences in local edges, Indicates the first The first component The length of a local edge, Indicates the first The first component The length of a local edge, Indicates the first The maximum value among the lengths of all local edges of a component. Indicates the first The first component The tilt angle of a local edge. Indicates the first The first component The tilt angle of a local edge. Indicates the first The maximum value of the tilt angles of all local edges of a component. Indicates the first The number of local edges of each component Represents the absolute value function. express function, express function, It is an exponential function with the natural constant as its base.

2. The method for identifying surface defects of integrated circuit board chips based on computer vision according to claim 1, characterized in that, The specific steps for obtaining several clusters are as follows: Using the elbow method, the optimal number of clusters is obtained when clustering all pixels in the chip package image using the K-means clustering algorithm. Let the number of clusters be Based on the grayscale value of each pixel in the chip package image, the K-means clustering algorithm is used to cluster all pixels in the chip package image to obtain... There are several clusters.

3. The method for identifying surface defects of integrated circuit board chips based on computer vision according to claim 1, characterized in that, The specific steps for obtaining the region within each component and the edge of each component are as follows: According to the first The pixels within a component are used to obtain several regions, which are denoted as the nth region. The area within each component; the first The first component The edge of the i-th region is denoted as the i-th region. The edge of each component.

4. The method for identifying surface defects of integrated circuit board chips based on computer vision according to claim 1, characterized in that, The specific steps for dividing the edge of each component into several local edges of each component are as follows: Through... Corner detection algorithms acquire all corner points within the chip package image, obtaining the first... The corner points on the edge of the component; the first The corner points on the edges of each component are used as segmentation points to divide the first component into segments. Each edge of a component is divided into multiple segments, and each segment is denoted as the first segment. Local edges of individual components.

5. The method for identifying surface defects of integrated circuit board chips based on computer vision according to claim 1, characterized in that, The specific steps for obtaining the length and tilt angle of the local edge of each component are as follows: Obtain the first... The first component The 8-adjacent chain code of a local edge; if the first The first component If the most frequent number in the 8-adjacent chain code of a local edge is less than 4, then the 1st... The first component The angle corresponding to the most frequent number in the 8-adjacent chain codes of a local edge is denoted as the i-th. The first component The tilt angle of the local edge; if the first The first component If the number that appears most frequently in the 8-adjacent chain codes of a local edge is greater than or equal to 4, then the 1st... The first component The angle corresponding to the value obtained by subtracting 4 from the most frequent number in the 8-adjacency chain code of a local edge is denoted as . The first component The tilt angle of the local edge; the first The first component The number of pixels contained on a local edge is denoted as the i-th. The first component The length of a local edge.

6. The method for identifying surface defects of integrated circuit board chips based on computer vision according to claim 1, characterized in that, The specific steps for dividing all regions within a component into shape defect regions and normal shape regions are as follows: Preset a threshold for the degree of defect manifestation. ;like Then the first The first component The region containing a local edge is denoted as the shape defect region; if Then the first The first component The region containing a local edge is denoted as the normally shaped region. Indicates the first The first component The degree of defect manifestation at a local edge.

7. The method for identifying surface defects of integrated circuit board chips based on computer vision according to claim 1, characterized in that, The specific steps for obtaining the region to be analyzed within each component are as follows: If the first... If the number of normally shaped areas within a component is 1, then... The normally shaped area within a component is divided into several identically sized... A rectangular region, denoted as the first... The region to be analyzed within each component; among which... The width of the preset rectangular area; if the first If the number of normally shaped regions within a component is greater than 1, then the first component will be... The normally shaped area within a component is denoted as the [number]th [component]. The area to be analyzed within each component.

8. The method for identifying surface defects of integrated circuit board chips based on computer vision according to claim 1, characterized in that, The specific calculation formula for obtaining the surface defect degree of each region to be analyzed within each component is as follows: In the formula, Indicates the first Within the first component The degree of surface defects in the area to be analyzed Indicates the first Within the first component The number of pixels contained in the region to be analyzed. Indicates the first Within the first component The first region to be analyzed The grayscale value of each pixel Indicates the first Within each component, in all regions to be analyzed, the first... The average grayscale value of each pixel Indicates the first The number of regions to be analyzed within each component. Indicates the first Within the first component The information entropy of the grayscale values ​​of pixels in the region to be analyzed. Indicates the first Within the first component The information entropy of the grayscale values ​​of pixels in the region to be analyzed. Indicates the first Within the first component The region to be analyzed and the first The absolute value of the difference in the number of pixels in the region to be analyzed. To prevent hyperparameters with a denominator of 0, Indicates the first The variance of the information entropy of the grayscale values ​​of pixels in all regions to be analyzed within a component. Represents the absolute value function. express function, express function.

9. The method for identifying surface defects of integrated circuit board chips based on computer vision according to claim 1, characterized in that, The specific steps for obtaining the region containing surface defects are as follows: Preset a surface defect severity threshold. ,like Then the first Within the first component The region to be analyzed is denoted as the region containing surface defects. Indicates the first Within the first component The degree of surface defects in the area to be analyzed.

Images