Image recognition-based electronic component defect detection method and system

Abstract

The present application relates to the technical field of electronic component image feature detection, and particularly relates to an electronic component defect detection method and system based on image recognition. An electronic component image is obtained and uniformly divided into multiple rectangular regions. A reference region is selected, and an information quantity descriptor is calculated according to the pixel neighborhood gray difference thereof. All regions are clustered and a feature region is selected. Feature coefficients are generated by combining the information quantity descriptor difference of adjacent regions in the feature region and the pixel gray and gradient difference in the AKAZE multi-scale space. The coefficients are used to register the to-be-detected image and the template image, and finally the accurate defect detection of the electronic component is realized. The present application can obtain the significant feature points in the electronic component image, realize the accurate registration of the electronic component image and the template image, and further accurately identify the defects of the electronic component.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electronic component image feature detection, and particularly relates to an electronic component defect detection method and system based on image recognition. BACKGROUND

[0002] In the integrated circuit manufacturing process, surface defect detection of integrated circuits is a key step to ensure the reliability and performance of electronic devices. Integrated circuits, as a kind of electronic component, have a complex manufacturing process involving multiple fine processes. Through surface defect detection, defective chips can be found and removed in time, thereby improving the yield of products. Template matching can quickly and accurately identify the regions inconsistent with the template by comparing the image of the region to be detected with the known standard template without defects, so as to determine whether there is a defect. Due to the slight changes in the manufacturing process, such as slight displacement, rotation or scaling of the wafer, there may be geometric deviations between the image to be detected and the template. If direct matching is performed, these deviations may be misjudged as defects, thereby reducing the accuracy of detection.

[0003] In the prior art, the feature points in the image are often detected by using algorithm, but the construction of the scale space and the detection of the feature points in the traditional algorithm are usually performed in the global range of the image. The feature points found in the global image may not always be representative, especially in some high-noise or low-texture regions. The algorithm may detect some unstable or insignificant feature points, thereby affecting the registration accuracy. SUMMARY

[0004] In order to solve the technical problem that in the traditional feature point detection algorithm, the construction of the scale space and the detection of the feature points are usually performed in the global range of the image, the feature points found in the global image may not always be representative, the algorithm may detect some unstable or insignificant feature points, thereby affecting the registration accuracy, and thus the electronic component defects cannot be accurately detected, the purpose of the present application is to provide an electronic component defect detection method and system based on image recognition, and the technical scheme adopted is as follows: an electronic component defect detection method based on image recognition, the method comprises: acquiring an electronic component image; dividing the electronic component image into multiple rectangular regions of a preset size; selecting one rectangular region as a reference region; obtaining an information descriptor of the reference region according to the gray difference between each pixel point in the reference region and each other pixel point in the preset neighborhood; clustering all rectangular regions according to the information descriptor of the reference region, to obtain all clustering clusters; screening all clustering clusters according to the information descriptor of all rectangular regions in each clustering cluster, to obtain all feature regions; and according to the information descriptor difference between adjacent rectangular regions in the feature region, and the gray difference between the pixel points in the feature region and The gray scale difference and the gradient difference between the same pixel points in different scale spaces under the algorithm are obtained to obtain feature coefficients of the feature region; the feature coefficients are used to register the electronic component image and a template electronic component image; and the electronic component is detected for defects according to the registered electronic component image.

[0005] Further, the method for obtaining the information amount descriptor comprises: obtaining the information amount descriptor according to an information amount descriptor calculation formula, and the information amount descriptor calculation formula is as follows: In the formula, represents the information amount descriptor of the reference region; represents the number of pixel points in the reference region; represents the number of other pixel points in a preset neighborhood of the i-th pixel point in the reference region; represents the number of other pixel points in a preset neighborhood of the i-th pixel point in the reference region; represents the gray scale value of the i-th pixel point in the reference region; represents the gray scale value of the i-th pixel point in the reference region; represents the gray scale value of the i-th other pixel point in a preset neighborhood of the i-th pixel point in the reference region; represents the gray scale value of the i-th other pixel point in a preset neighborhood of the i-th pixel point in the reference region; represents the gray scale value of the i-th other pixel point in a preset neighborhood of the i-th pixel point in the reference region; represents an absolute value function.

[0006] Further, the method for obtaining the feature region comprises: clustering all the rectangular regions by using the information amount descriptors of the rectangular regions to obtain all feature clustering clusters; taking a total region composed of rectangular regions in a cluster with the maximum mean value of the information amount descriptors as a to-be-analyzed region; and marking the continuous rectangular regions in the to-be-analyzed region to take a region composed of the continuous rectangular regions as each feature region.

[0007] Further, the method for obtaining the feature coefficient comprises: obtaining the feature coefficient according to a feature coefficient calculation formula, and the feature coefficient calculation formula is as follows: In the formula, represents the feature coefficient of each feature region; represents the number of rectangular regions contained in each feature region; represents the information amount descriptor of the i-th rectangular region in each feature region; represents the information amount descriptor of the i-th rectangular region in each feature region; represents the information amount descriptor of the i-th rectangular region in each feature region; represents the information amount descriptor of the i-th rectangular region in each feature region; represents the number of pixel points in each feature region; represents the pixel point serial number in each feature region; represents the number of scale spaces under the algorithm; represents Scale space index under the algorithm; Represents the first [characteristic] region within each feature region. The grayscale value of each pixel; express The first under the algorithm In the i-th scale space, the first The grayscale value of each pixel; Represents the first [characteristic] region within each feature region. Gradient values ​​of each pixel; express The first under the algorithm In the i-th scale space, the first Gradient values ​​of each pixel; This represents the absolute value function.

[0008] Further, using the feature coefficients, the electronic component image is registered with the template electronic component image, including: sorting the feature regions according to the feature coefficients from high to low to obtain a feature region sequence; in In the algorithm, feature point detection is performed on each feature region according to the order of the feature regions in the feature region sequence to obtain all feature points of each feature region; a template image of the electronic component is obtained as a template electronic component image; and the feature regions in the electronic component image are registered with the template electronic component image in sequence according to the order of the feature regions in the feature region sequence.

[0009] Further, defect detection of electronic components is performed on the registered electronic component images, including: calculating the residual between each corresponding pixel point of the registered electronic component image and the template electronic component image as the first residual, and calculating the mean of all first residuals as the global residual; when the difference between the first residual and the global residual is greater than a preset second threshold, the pixel point corresponding to the first residual is marked as a defect pixel point; and traversing all pixels in the registered electronic component image to obtain the location of all defect pixel points.

[0010] An image recognition-based electronic component defect detection system is provided. The system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the image recognition-based electronic component defect detection method described above.

[0011] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the image recognition-based electronic component defect detection method described above.

[0012] The computer device comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the image recognition-based electronic component defect detection method.

[0013] The present application has the following advantages: the present application obtains an electronic component image of a to-be-tested component, and analyzes the electronic component image based on the algorithm; since the algorithm defines feature points in different scale spaces of the image, and the positions of the feature points are in local regions, the electronic component image is evenly divided into a plurality of rectangular regions of a preset size; the physical structure information of integrated circuits contained in different rectangular regions is different, for a wire, the width, interval, or complexity of arrangement of the wire will be obviously different in different local regions, and different pads will also have different sharp degrees due to different shapes, which will also affect the clarity of the texture, therefore, according to the gray level difference between each pixel point in the reference region and each other pixel point in the preset neighborhood, an information quantity descriptor of the reference region is obtained, and the image information in each rectangular region is analyzed through the information quantity descriptor; the information quantity descriptor is taken as the attribute of each rectangular region, so that the most characteristic region in the electronic component image is marked by using the similarity of the information quantities of different rectangular regions, which is beneficial to the selection of subsequent feature points, therefore, all the rectangular regions are clustered according to the information quantity descriptor of the reference region, and all the clustering clusters are obtained; all the clustering clusters are screened according to the information quantity descriptors of all the rectangular regions in each clustering cluster, and all the feature regions are obtained; since the algorithm needs to find feature points in local regions, the above step reduces the traversal range of the algorithm, and subsequent operations need to find feature points with more obvious anti-interference ability in a smaller traversal range, therefore, a feature coefficient in the feature region is defined, the anti-interference ability of the feature region is reflected through the feature coefficient, and the stronger the anti-interference ability is, the more obvious feature points the feature region is likely to have, so the feature coefficient of the feature region is obtained according to the information quantity descriptor difference between adjacent rectangular regions in the feature region and the gray level difference and gradient difference between the same pixel points in different scale spaces under the algorithm; then, a template image of the electronic component obtained by using the prior art is used to register the electronic component image; and the electronic component is detected for defects according to the registered electronic component image. The present application can obtain significant feature points in the electronic component image, realize accurate registration of the electronic component image and the template image, and then accurately identify the defects of the electronic component. BRIEF DESCRIPTION OF DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, and the advantages thereof, below will briefly introduce the drawings needed in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative effort.

[0015] Figure 1 A flow chart of an electronic component defect detection method based on image recognition provided by an embodiment of the present application; Figure 2 A block diagram of an electronic component defect detection system based on image recognition provided by an embodiment of the present application. DETAILED DESCRIPTION

[0016] In order to further illustrate the technical means and effects taken by the present application to achieve the predetermined purposes, below will combine the drawings and preferred embodiments to specifically describe the electronic component defect detection method and system based on image recognition according to the present application, the specific implementation, structure, features and effects thereof, in detail as follows. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0017] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0018] Below will specifically describe the specific scheme of the electronic component defect detection method and system based on image recognition provided by the present application in combination with the drawings.

[0019] Please refer to Figure 1 which shows an electronic component defect detection method based on image recognition provided by an embodiment of the present application, the method comprising: step S1: acquiring an electronic component image.

[0020] The embodiment of the present application is mainly applied to the accurate registration scene of electronic component images and template images, so the electronic component image to be detected is first collected.

[0021] In an embodiment of the present application, the integrated circuit sample to be detected is stably installed on the inspection table, ensuring that its surface is flat and can be clearly observed. The position and angle of the sample are adjusted according to the detection requirements, ensuring that the region to be detected is in the field of view of the camera. Then, a high-resolution industrial camera is used to collect images on the selected region to be detected. The camera is a black and white industrial camera, which directly collects a single-channel grayscale image. The grayscale image is denoised to obtain the electronic component image required in the subsequent steps.

[0022] It should be noted that, since The algorithm has strong robustness for images with complex textures, and is suitable for application in defect detection of integrated circuits by templates, so in the embodiment of the present application, subsequent operations are based on Algorithm analysis.

[0023] Step S2: divide the electronic component image into a plurality of rectangular regions of a preset size; optionally, one rectangular region is taken as a reference region; obtain the information descriptor of the reference region according to the gray difference between each pixel point in the reference region and each other pixel point in the preset neighborhood; cluster all rectangular regions according to the information descriptor of the reference region, to obtain all clustering clusters; filter all clustering clusters according to the information descriptor of all rectangular regions in each clustering cluster, to obtain all feature regions; obtain the feature coefficient of the feature region according to the information descriptor difference between adjacent rectangular regions in the feature region, and the gray difference and gradient difference between the same pixel points in different scale spaces under the Algorithm.

[0024] Since The algorithm defines feature points in different scale spaces of the image, and the positions of the feature points are in local regions, so in the embodiment of the present application, the electronic component image is divided into a plurality of rectangular regions of a preset size.

[0025] In an embodiment of the present application, the preset size is set to Rectangular region, and the electronic component image is divided into a plurality of rectangular regions of the same preset size, and the number of rectangular regions is determined by the size of the electronic component image, which is not limited here.

[0026] Different rectangular regions contain different physical structure information of integrated circuits. For wires, the width, spacing, or complexity of arrangement of the wires will show obvious differences in different local regions, and different pads will also affect the clarity of the texture due to different sharpness caused by different shapes, so in the embodiment of the present application, the information descriptor of the reference region is obtained according to the gray difference between each pixel point in the reference region and each other pixel point in the preset neighborhood, and the image information in each rectangular region is analyzed through the information descriptor.

[0027] Preferably, in an embodiment of the present application, the information descriptor acquisition method comprises: obtaining the information descriptor according to the information descriptor calculation formula, and the information descriptor calculation formula is as follows: In the formula, Indicates the information descriptor of the reference region. represents the number of pixel points in the reference region; represents the number of other pixel points in the preset neighborhood of the pixel point in the reference region; represents the gray value of the pixel point in the reference region; represents the gray value of the pixel point in the preset neighborhood of the pixel point in the reference region; represents the absolute value function.

[0028] In the information quantity descriptor calculation formula, the greater the gray difference between each pixel point in the rectangular region and other pixel points in the preset neighborhood , the more likely the pixel point is at the connection between the pad edge and the wire, and the more such pixel points are distributed in the rectangular region, the greater the information quantity contained in the rectangular region, and the greater the information quantity descriptor; , the greater the gray contrast between the pixel point and other pixel points in the preset neighborhood, which means that the pixel point is more likely to be at the connection between the pad edge and the wire, and the more such pixel points are distributed in the rectangular region, the greater the information quantity contained in the rectangular region, and the greater the information quantity descriptor; and the degree of gray contrast between the pixel point and other pixel points in the preset neighborhood is amplified through the exponential function , so as to increase the significance of the information quantity description in the rectangular region; the same analysis is performed on each pixel point to obtain the information quantity descriptor of the rectangular region.

[0029] The information quantity descriptor is taken as the attribute of each rectangular region, so that the most characteristic region in the electronic component image is marked by using the similarity of the information quantity between different rectangular regions, thereby facilitating the selection of the feature point subsequently. Therefore, in the embodiment of the present application, all the rectangular regions are clustered according to the information quantity descriptor of the reference region to obtain all the clustering clusters; and all the clustering clusters are screened according to the information quantity descriptor of all the rectangular regions in each clustering cluster to obtain all the feature regions.

[0030] Preferably, in an embodiment of the present application, the feature region acquisition method comprises: clustering all the rectangular regions by using the information quantity descriptor of each rectangular region to obtain all the feature clustering clusters; in an embodiment of the present application, a clustering algorithm is used for operation, and the specific steps are technical means familiar to those skilled in the art, which will not be described here.

[0031] ​The total region consisting of the rectangular regions contained in the cluster with the largest mean information content descriptor in the feature cluster is taken as the region to be analyzed. Since the different rectangular regions in the region to be analyzed are discretely distributed and are formed by a large area composed of multiple continuous rectangular regions, the continuous rectangular regions are marked in the region to be analyzed, and the region composed of the continuous rectangular regions is taken as each feature region.

[0032] because The algorithm needs to find feature points in a local region. The steps described above narrow the algorithm's traversal range. Subsequent operations need to find feature points with strong anti-interference capabilities within a smaller traversal range. Therefore, feature coefficients are defined for the feature region to reflect its anti-interference capability. Stronger anti-interference capability indicates that the feature region may contain more obvious feature points. Thus, in this embodiment, based on the information content descriptor difference between adjacent rectangular regions in the feature region, and the pixel values ​​within the feature region... The algorithm calculates the grayscale and gradient differences between the same pixels at different scales to obtain the feature coefficients of the feature region.

[0033] Preferably, in one embodiment of the present invention, the method for obtaining the feature coefficients includes: obtaining the feature coefficients according to the feature coefficient calculation formula, the feature coefficient calculation formula being as follows: In the formula, Represents the feature coefficients of each feature region; This indicates the number of rectangular regions contained within each feature region; Represents the first [characteristic] region within each feature region. Information descriptor for a rectangular region; Represents the first [characteristic] region within each feature region. Information descriptor for a rectangular region; This indicates the number of pixels within each feature region; This represents the pixel number within each feature region; express Number of scale spaces under the algorithm; express Scale space index under the algorithm; Represents the first [characteristic] region within each feature region. The grayscale value of each pixel; express The first under the algorithm In the i-th scale space, the first The grayscale value of each pixel; Represents the first [characteristic] region within each feature region. Gradient values ​​of each pixel; express The first under the algorithm Gradient value of a pixel point in a scale space represents an absolute value function. represents an absolute value function.

[0034] In the feature coefficient calculation formula, the information quantity descriptor difference mean between every two adjacent rectangular regions in the feature region is smaller, which indicates that there is a more complex physical structure in the different rectangular regions in the feature region, and the probability of obtaining a feature point with a higher confidence degree in the feature region is greater, and therefore the corresponding feature coefficient is greater. The algorithm can perform a convolution operation on the feature region through a Gaussian filter. As the filter kernel increases, the blurring degree of the feature region gradually deepens. The filtered result obtained by filtering the feature region each time is a scale space. The larger the filter kernel, the more obvious the blurring degree in the corresponding scale space, and therefore the smaller the filter kernel, the lower the corresponding scale space. The filter kernel size is an odd number, starting from to . The feature point under a low scale appears as a local extreme point such as a corner point, a metal line connection point, etc. The feature point under a high scale appears as some structure such as a metal interconnection layer with a more obvious volume, a transistor overall structure, etc. Therefore, the greater the gray difference and gradient difference between the pixel point under the original scale and the same pixel point under different scale spaces in the feature region, the more the number of feature points in the feature region, and therefore the greater the corresponding feature coefficient. and are weighted and averaged to form the feature coefficient of each feature region in combination with .

[0035] In an embodiment of the present application, the rectangular regions in the feature region are sorted in a top-down and left-right order to obtain the calculation order of the different rectangular regions in each feature region.

[0036] The template image of the electronic component obtained by using the prior art is used to register the electronic component image. Preferably, in an embodiment of the present application, the feature coefficient is used to register the electronic component image and the template electronic component image, including: sorting the feature regions in a descending order of the feature coefficients to obtain a feature region sequence; and detecting feature points in each feature region according to the order of the feature regions in the feature region sequence to obtain all the feature points of each feature region. In the algorithm, feature points are detected in each feature region according to the order of the feature regions in the feature region sequence to obtain all the feature points of each feature region. It should be noted that the feature point obtaining step is a technical means known to those skilled in the art, and will not be described here.

[0037] obtaining a template image of the electronic component as a template electronic component image; and sequentially registering the feature regions in the electronic component image with the template electronic component image according to the sequence of the feature regions. It should be noted that the registration process is a technical means known to those skilled in the art, and will not be described here.

[0038] Step S3: performing defect detection on the electronic component according to the registered electronic component image.

[0039] Preferably, in an embodiment of the present application, the defect detection on the electronic component according to the registered electronic component image comprises: calculating a residual error between each corresponding pixel point of the registered electronic component image and the template electronic component image as a first residual error, and calculating a mean value of all the first residual errors as a global residual error.

[0040] When the difference between the first residual error and the global residual error is greater than a preset second threshold value, the pixel point corresponding to the first residual error is marked as a defect pixel point; and all the defect pixel points are obtained by traversing all the pixel points in the registered electronic component image. In an embodiment of the present application, the preset second threshold value is set to 30%. It should be noted that the preset second threshold value can be set by the user and is not limited herein.

[0041] In summary, an electronic component image is obtained; the electronic component image is divided into a plurality of rectangular regions of a preset size; one of the rectangular regions is selected as a reference region; an information descriptor of the reference region is obtained according to the gray level difference between each pixel point in the reference region and each other pixel point in a preset neighborhood; all the rectangular regions are clustered according to the information descriptors of the rectangular regions in each cluster, and all the clusters are obtained; all the clusters are screened according to the information descriptors of all the rectangular regions in each cluster, and all the feature regions are obtained; feature coefficients of the feature regions are obtained according to the information descriptor difference between adjacent rectangular regions in the feature regions, and the gray level difference and gradient difference between the same pixel points in different scale spaces under the algorithm; the electronic component image and the template electronic component image are registered by using the feature coefficients; and defect detection is performed on the electronic component according to the registered electronic component image. In summary, an electronic component image is obtained; the electronic component image is divided into a plurality of rectangular regions of a preset size; one of the rectangular regions is selected as a reference region; an information descriptor of the reference region is obtained according to the gray level difference between each pixel point in the reference region and each other pixel point in a preset neighborhood; all the rectangular regions are clustered according to the information descriptors of the rectangular regions in each cluster, and all the clusters are obtained; all the clusters are screened according to the information descriptors of all the rectangular regions in each cluster, and all the feature regions are obtained; feature coefficients of the feature regions are obtained according to the information descriptor difference between adjacent rectangular regions in the feature regions, and the gray level difference and gradient difference between the same pixel points in different scale spaces under the algorithm; the electronic component image and the template electronic component image are registered by using the feature coefficients; and defect detection is performed on the electronic component according to the registered electronic component image.

[0042] An embodiment of the present application provides an electronic component defect detection system based on image recognition, which comprises: Figure 2As shown, the system comprises a memory, a processor and a computer program, wherein the memory is used to store the corresponding computer program, the processor is used to run the corresponding computer program, and the computer program can realize the method described in steps S1-S3 when running in the processor, and specifically comprises: an image acquisition module 101, used to acquire an electronic component image; an image analysis module 102, used to divide the electronic component image into a plurality of rectangular regions of a preset size; optionally, one rectangular region is used as a reference region; according to the gray difference between each pixel point in the reference region and each other pixel point in a preset neighborhood, an information quantity descriptor of the reference region is obtained; according to the information quantity descriptor of the reference region, all rectangular regions are clustered to obtain all clustering clusters; according to the information quantity descriptor of all rectangular regions in each clustering cluster, all clustering clusters are screened to obtain all feature regions; according to the information quantity descriptor difference between adjacent rectangular regions in the feature region, and the gray difference and gradient difference between the same pixel points in different scale spaces under the algorithm, a feature coefficient of the feature region is obtained; the electronic component image and the template electronic component image are registered by using the feature coefficient; and a defect detection module 103, used to detect defects of the electronic component according to the registered electronic component image.

[0043] A third object of the embodiment of the present application is to provide a computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor realizes the method described in steps S1-S3 when executing the computer program.

[0044] A fourth object of the embodiment of the present application is to provide a computer readable storage medium, which stores a computer program, and the computer program realizes the method described in steps S1-S3 when executed by a processor.

[0045] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0046] Each embodiment in the specification is described in a progressive manner, and the same and similar parts between each embodiment can be referred to each other, and each embodiment mainly describes the difference from other embodiments.​

Claims

1. An electronic component defect detection method based on image recognition, characterized by, The method comprises: acquiring an electronic component image; dividing the electronic component image into a plurality of rectangular regions of a preset size; optionally taking one rectangular region as a reference region; obtaining an information descriptor of the reference region according to the gray level difference between each pixel point in the reference region and each other pixel point in a preset neighborhood; clustering all rectangular regions according to the information descriptor of the reference region to obtain all clustering clusters; screening all clustering clusters according to the information descriptor of all rectangular regions in each clustering cluster to obtain all feature regions; obtaining a feature coefficient of the feature region according to the information descriptor difference between adjacent rectangular regions in the feature region and the gray level difference and gradient difference between the same pixel points in different scale spaces under the algorithm; using the feature coefficient to register the electronic component image with a template electronic component image; and detecting defects of the electronic component according to the registered electronic component image. The method for obtaining the information descriptor comprises: obtaining the information descriptor according to an information descriptor calculation formula, as shown below: In the formula, represents the information descriptor of the reference region; represents the number of pixel points in the reference region; represents the number of other pixel points in a preset neighborhood of the i-th pixel point in the reference region; represents the gray level value of the i-th pixel point in the reference region; represents the gray level value of the i-th pixel point in the reference region; represents the gray level value of the i-th pixel point in the reference region; represents the gray level value of the i-th pixel point in the reference region; represents the gray level value of the i-th pixel point in the reference region; represents the gray level value of the i-th pixel point in the reference region; represents an absolute value function; the method for obtaining the feature coefficient comprises: obtaining the feature coefficient according to a feature coefficient calculation formula, as shown below: In the formula, represents the feature coefficient of each feature region; represents the number of rectangular regions contained in each feature region; represents the information descriptor of the i-th rectangular region in each feature region; represents the information descriptor of the i-th rectangular region in each feature region; represents the information descriptor of the i-th rectangular region in each feature region; represents the number of pixel points in each feature region; represents the pixel point serial number in each feature region; represents represents the number of scale spaces under the algorithm; represents the number of scale spaces under the algorithm; represents the number of scale spaces under the algorithm; Scale space index under algorithm; represents the gray value of the i-th pixel point in the j-th scale space under algorithm; represents the gradient value of the i-th pixel point in the j-th scale space under algorithm; represents the gradient value of the i-th pixel point in the j-th scale space under algorithm; represents the gradient value of the i-th pixel point in the j-th scale space under algorithm; represents the gradient value of the i-th pixel point in the j-th scale space under algorithm; represents the absolute value function.​​​​​​​ 2. The electronic component defect detection method based on image recognition according to claim 1, characterized in that, The feature region acquisition method comprises: clustering all the rectangular regions by using the information quantity descriptor of each rectangular region to obtain all feature clustering clusters; taking a total region composed of rectangular regions contained in a cluster with the maximum mean information quantity descriptor in the feature clustering cluster as a region to be analyzed; and marking continuous rectangular regions in the region to be analyzed and taking a region composed of the continuous rectangular regions as each feature region.

3. The electronic component defect detection method based on image recognition according to claim 1, characterized in that, The electronic component image is registered with a template electronic component image by using the characteristic coefficients, including: arranging the characteristic regions according to the characteristic coefficients from high to low to obtain a characteristic region sequence; and In the algorithm, the feature point detection is performed on each characteristic region according to the sequence of the characteristic regions in the characteristic region sequence to obtain all feature points of each characteristic region; a template image of the electronic component is obtained as a template electronic component image; and the characteristic regions in the electronic component image are sequentially registered with the template electronic component image according to the sequence of the characteristic regions in the characteristic region sequence.

4. The electronic component defect detection method based on image recognition according to claim 1, characterized in that, According to the registered electronic component image, the electronic component is detected for defects, comprising: calculating a residual error between each corresponding pixel point of the registered electronic component image and the template electronic component image as a first residual error, and calculating a mean value of all the first residual errors as a global residual error; when a difference between the first residual error and the global residual error is greater than a preset second threshold value, taking the pixel point corresponding to the first residual error as a defect pixel point and marking the defect pixel point; and traversing all the pixel points in the registered electronic component image to obtain positions of all the defect pixel points.

5. An image recognition-based electronic component defect detection system, the system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program to realize the steps of the image recognition-based electronic component defect detection method according to any one of claims 1 to 4.

6. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 5. The computer program is executed by the processor to realize the steps of the image recognition-based electronic component defect detection method according to any one of claims 1 to 4.

7. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program to realize the steps of the image recognition-based electronic component defect detection method according to any one of claims 1 to 4.

Images