Solar cell appearance defect detection method, device and equipment based on machine vision and storage medium

Abstract

The invention provides a solar cell appearance defect detection method, device and equipment based on machine vision and a storage medium. The detection method comprises the following steps: acquiring a whole image containing a battery piece; selecting a whole image of a sample without defects from all the whole images containing the battery piece, and generating a battery piece area template image of the sample; obtaining a whole image of a to-be-measured image, and generating a battery piece region rectification image of the to-be-measured image; carrying out mutual difference operation on the battery piece area rectification image of the to-be-measured image and the battery piece area template image of the sample to obtain a mutual difference operation result; and performing defect classification according to a mutual difference operation result, and filtering the defects according to a set judgment rule to determine whether the battery piece is qualified or not. According to the invention, various defects of the battery piece can be rapidly and effectively detected, and the method can adapt to the change of the model.

Description

Technical Field

[0001] The present invention relates to the technical field of solar cell chips, and in particular to a method, device, equipment and storage medium for detecting appearance defects of solar cell chips based on machine vision. Background Art

[0002] Nowadays, machine vision-based solar cell appearance defect detection has become the mainstream method for cell defect detection and has become a huge driving force for the construction of smart workshops. However, the existing appearance defect detection methods have some shortcomings, especially in terms of version switching and detection efficiency.

[0003] The appearance defects of the cell mainly include grid line defects of the cell, such as broken grid, bright node, electrode abnormality, and defects such as foreign matter (dirt) and leakage of slurry in the background. In particular, grid line defects are highly dependent on the pattern, which brings great difficulty to the detection of grid line defects.

[0004] In recent years, in order to solve the above-mentioned problems of low detection efficiency and multiple grid line patterns, the industry has done a lot of research work. The mainstream method is to encapsulate the grid line extraction function into finer and more universal units. When a new pattern appears, the parameters of the encapsulated grid line extraction unit are fine-tuned to form a new grid line extraction method to adapt to different patterns.

[0005] With the advent of the new energy era, intelligent detection systems equipped with visual systems are being promoted on a large scale with their rich functions and precise detection effects. At present, photovoltaic cell workshops are also shifting from traditional manufacturing workshops to digital intelligent manufacturing workshops. While bringing huge market opportunities, it also brings great challenges to intelligent defect detection based on machine vision. Detection stability, accuracy, and efficiency have become the focus of increasing attention of users. Therefore, intelligent detection methods based on vision will inevitably become a new research direction and field.

[0006] Prior art 1, Chinese patent application CN109084957A proposes a defect detection and color sorting method for photovoltaic solar cells. The method first collects silicon wafer images and performs preprocessing to achieve the purpose of color calibration; secondly, the silicon wafer is positioned; on this basis, damage detection, grid line extraction, grid line detection, main grid detection, and color spot detection are performed. The method also discloses a color sorting method to achieve color grading of solar mirror cabinet cells. This invention application proposes a defect detection and color sorting method for photovoltaic solar cells, which can achieve contactless detection and has the characteristics of high detection accuracy and good stability. The above method is a common method for defect detection of photovoltaic solar cells. Although it can achieve the purpose of defect detection, this method cannot fully adapt to all versions. Moreover, the invention application requires the extraction of grid lines and pixel-by-pixel analysis of the grid lines, and the detection efficiency is low.

[0007] Prior art 2, Chinese patent application CN116758080A proposes a method for detecting screen printing defects of solar cells. The invention first obtains the cell image, and then performs grayscale processing, filtering and noise reduction processing, and brightness equalization on the cell; then the image to be tested is registered, and the registered image is bilinearly interpolated and the neural network is trained to output the defects of the cell. Although the prior art 2 can quickly achieve the purpose of pattern adjustment and realize efficient defect detection, this method requires manual construction of a large number of data sets to support model training. At the same time, on the inference end, deep learning requires expensive computing resources to support. In addition, the trained model cannot provide effective parameters for users to quickly adjust the model, and the ability to adjust defects is poor, so it is greatly limited in use. Summary of the invention

[0008] In order to solve at least one technical problem in the prior art, the embodiments of the present invention provide a method, device, equipment and storage medium for detecting appearance defects of solar cells based on machine vision, which can quickly and effectively detect various defects of solar cells, and can adapt to changes in the format and reduce the cost of format replacement. In order to achieve the above technical objectives, the technical solution adopted by the embodiments of the present invention is:

[0009] In a first aspect, an embodiment of the present invention provides a method for detecting appearance defects of a solar cell based on machine vision, comprising the following steps:

[0010] Acquire the entire image including the battery cell;

[0011] Select a sample whole image without defects from all the whole images containing the battery cells, and generate a template image of the battery cell area of ​​the sample;

[0012] Obtain a whole image of the image to be tested, and generate a corrected image of the cell area of ​​the image to be tested;

[0013] Performing a mutual difference operation on the battery cell area correction image of the image to be tested and the battery cell area template image of the sample to obtain a mutual difference operation result;

[0014] Defects are classified according to the results of the mutual difference operation, and the defects are filtered according to the set judgment rules to determine whether the battery cell is qualified.

[0015] Furthermore, after obtaining the entire image including the battery cell, the following steps are also included:

[0016] Perform noise reduction on the entire image including the battery cell.

[0017] Furthermore, the step of selecting a sample whole image without defects from all the whole images containing the battery cells and generating a sample battery cell area template image specifically includes:

[0018] Select a whole image of a sample without defects;

[0019] Mark the conveyor belt area in the entire sample image and save the conveyor belt position coordinates;

[0020] Binarize the entire sample image to obtain a binarized image of the entire sample; and save the binarization threshold when the image is binarized;

[0021] Subtract the conveyor belt area from the entire binary image of the sample to obtain a difference area image of the sample, and perform a denoising operation on the difference area image of the sample; and record the denoising parameters in the denoising operation;

[0022] The difference area image of the sample after denoising is transformed into convexity to obtain the battery cell area of ​​the sample; the area image where the battery cell area of ​​the sample is located is intercepted from the entire sample image to obtain the sample battery cell area image after removing the conveyor belt; the center point and deflection angle of the sample battery cell area image are obtained by image processing of the sample battery cell area image;

[0023] Calculate a first affine transformation matrix according to the center point and deflection angle of the sample cell area image and the standard center point and standard angle 0 of the camera field of view;

[0024] Performing an affine transformation on the sample cell region image according to the first affine transformation matrix to obtain a cell region template image and save the image;

[0025] Select at least two positioning kernel templates on the battery cell area template image and save the position coordinates of the positioning kernels.

[0026] Furthermore, the step of acquiring a whole image of a test image and generating a cell region-corrected image of the test image specifically includes:

[0027] Get the entire image of a test image;

[0028] Read the saved binarization threshold, binarize the entire image of the image to be tested, and obtain the entire binarized image of the image to be tested;

[0029] Read the saved conveyor belt position coordinates, and generate the conveyor belt area according to the conveyor belt position coordinates; subtract the conveyor belt area from the entire binary image of the image to be tested to obtain the difference area image of the image to be tested;

[0030] Read the saved denoising parameters and perform denoising operation on the difference area image of the image to be tested;

[0031] The difference area image of the image to be tested after denoising is subjected to convex transformation to obtain the cell area of ​​the image to be tested; the area image where the cell area of ​​the image to be tested is located is intercepted from the entire image of the image to be tested to obtain the cell area image of the image to be tested after removing the conveyor belt; the center point and deflection angle of the cell area image of the image to be tested are obtained by image processing of the cell area image of the image to be tested;

[0032] Calculate the second affine transformation matrix according to the center point and deflection angle of the cell area image of the image to be tested and the standard center point and standard angle 0 of the camera field of view;

[0033] Read the saved positioning core template and the position coordinates of the positioning core; perform affine transformation according to the second affine transformation matrix to obtain the position coordinates of the positioning core after the affine transformation, intercept an image near the position coordinates of the positioning core after the affine transformation on the battery cell area image of the image to be tested, and use the saved positioning core template to perform template matching on the intercepted images to obtain the position coordinates of the positioning core on the battery cell area image of the image to be tested;

[0034] Calculate the transmission transformation matrix according to the position coordinates of the positioning kernel on the cell region image of the image to be tested and the position coordinates of the positioning kernel on the cell region template image of the sample;

[0035] The cell area image of the image to be measured is subjected to transmission transformation according to the transmission transformation matrix to obtain a cell area correction image.

[0036] Furthermore, the denoising operation is implemented by opening the operation kernel as a rectangle; and the denoising parameters are the length and width of the operation kernel.

[0037] Furthermore, performing a mutual difference operation on the cell region rectified image of the image to be tested and the cell region template image of the sample to obtain a mutual difference operation result specifically includes:

[0038] Subtract the template image of the cell area of ​​the sample from the corrected image of the cell area of ​​the image to be tested, and obtain the bright defect reference image after difference;

[0039] The dark defect reference image is obtained by subtracting the deflection correction image of the cell area of ​​the image to be tested from the template image of the cell area of ​​the sample.

[0040] Furthermore, the defect classification according to the mutual difference operation result and the filtering of the defects according to the set judgment rules to determine whether the battery cell is qualified specifically include:

[0041] Binarize the bright defect reference image to obtain a bright defect reference binary image; traverse the connected domains therein, and for each connected domain:

[0042] If the connected domain is on the grid line, it is marked as a bright node defect;

[0043] If the connected domain is not on the grid line, it is marked as a foreign body defect;

[0044] If the connected domain is on the electrode, it is marked as an abnormal electrode defect;

[0045] Binarize the dark defect reference image to obtain a dark defect reference binary image; traverse the connected domains therein, and for each connected domain:

[0046] If the connected domain is on the gate line, it is marked as a broken gate defect;

[0047] If the connected domain is on the electrode, it is marked as an electrode missing defect;

[0048] Defects are filtered according to the set judgment rules to determine whether the battery cell is qualified.

[0049] In a second aspect, an embodiment of the present invention provides a device for detecting appearance defects of solar cells based on machine vision, comprising:

[0050] An image acquisition module, used to acquire the entire image including the battery cell;

[0051] A first area image generation module is used to select a sample whole image without defects from all the whole images containing the battery cells, and generate a sample battery cell area template image;

[0052] The second area image generation module is used to obtain a whole image of a test image and generate a cell area correction image of the test image;

[0053] An image mutual difference module is used to perform mutual difference operation on the battery cell area correction image of the image to be tested and the battery cell area template image of the sample to obtain a mutual difference operation result;

[0054] The image defect detection module is used to classify defects according to the results of mutual difference operation, and filter the defects according to the set judgment rules to determine whether the battery cell is qualified.

[0055] In a third aspect, an embodiment of the present invention provides an electronic device, including:

[0056] a memory storing a computer program;

[0057] The processor is used to run the computer program, and when the computer program is run, the steps of the method for detecting appearance defects of solar cells based on machine vision as described above are executed.

[0058] In a fourth aspect, an embodiment of the present invention provides a storage medium, wherein a computer program is stored in the storage medium, and the computer program is configured to execute the steps of the method for detecting appearance defects of solar cells based on machine vision as described above when running.

[0059] The beneficial effects of the technical solution provided by the embodiment of the present invention are as follows: according to the solar cell appearance defect detection method proposed in the present invention, under this method, the appearance defects of the cell, such as broken grid, bright node, electrode abnormality, foreign matter (dirt) and other defects can be detected quickly and effectively. While ensuring the accuracy of detection, this method takes into account the detection speed, and guarantees the efficient operation of the production line to the greatest extent, which can significantly reduce the detection rhythm and provide strong support for improving production capacity. At the same time, this method can quickly adapt to different versions according to the actual situation of the user, realize the switching detection of different versions, and effectively reduce the time loss brought by the version switching. Under this method, the appearance defects of the cell can be effectively detected, which can greatly reduce the defect rate of the cell and improve the photoelectric conversion rate and service life of the photovoltaic module. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 4 is a flow chart of the detection method in an embodiment of the present invention.

[0061] Figure 2 This is a flow chart of generating a sample battery cell area template image in an embodiment of the present invention.

[0062] Figure 3 The flowchart of the battery cell area deflection correction image for generating the image to be tested in the embodiment of the present invention.

[0063] Figure 4 Schematic diagram of a detection device in an embodiment of the present invention.

[0064] Figure 5 Schematic diagram of an electronic device in an embodiment of the present invention. DETAILED DESCRIPTION

[0065] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0066] In the description of the embodiments of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance.

[0067] In the description of the embodiments of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, it can also be the internal connection of two components, it can be a wireless connection, or it can be a wired connection. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0068] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0069] The embodiment of the present invention provides a method for detecting appearance defects of solar cells based on machine vision, comprising the following steps:

[0070] Step S10, acquiring the entire image including the battery cell;

[0071] In a specific embodiment, the battery cell is carried on a conveyor belt, and an industrial camera is used to obtain a whole image including the battery cell; when obtaining the image, four strip-shaped red light sources are used as the light source, which are respectively placed above the four sides of the battery cell and have a certain inclination angle; the light source is adjusted so that the grid lines in the whole image obtained are clear;

[0072] Since a whole image of a sample without defects needs to be selected, a plurality of whole images including the battery cells can be obtained in step S10;

[0073] Optionally, step S15, performing noise reduction processing on the entire image including the battery cell;

[0074] In a specific embodiment, abnormal noise in the entire image including the battery cell can be removed by Gaussian filtering to achieve image noise reduction processing;

[0075] Step S20, selecting a sample whole image without defects from all the whole images containing the battery cells, and generating a sample battery cell area template image; specifically comprising:

[0076] Step S201, selecting a defect-free sample whole image;

[0077] Step S202, marking the conveyor belt area in the entire sample image and saving the conveyor belt position coordinates;

[0078] In this embodiment, the conveyor belt area can be manually marked, denoted as Reg belt , save the conveyor belt position coordinates belt to the local; the conveyor belt is two parallel narrow belts;

[0079] Step S203, binarizing the entire sample image to obtain a binarized image of the entire sample; and saving the binarization threshold when binarizing the image;

[0080] In this embodiment, the entire binary image of the sample is denoted as Reg cell , the binarization threshold is recorded as thr; the selected binarization threshold should make the edge of the battery cell and the background clearly visible;

[0081] Step S204, subtracting the conveyor belt area from the entire binary image of the sample to obtain a difference area image of the sample, and performing a denoising operation on the difference area image of the sample; and recording the denoising parameters in the denoising operation;

[0082] In this embodiment, the difference region image is recorded as Reg diff ;

[0083] In this embodiment, the denoising operation is implemented by opening the operation kernel as a rectangle; the denoising parameters are the length and width of the operation kernel; the length and width of the operation kernel are denoted as l×w;

[0084] Step S205, performing convexity transformation on the difference area image of the denoised sample to obtain the cell area of ​​the sample; intercepting the area image where the cell area of ​​the sample is located from the entire sample image to obtain the sample cell area image after removing the conveyor belt; performing image processing on the sample cell area image to obtain the center point and deflection angle of the sample cell area image;

[0085] In this embodiment, the difference region image of the denoised sample is converted into a convexity to obtain the cell region of the sample, which is recorded as Reg exp ; Extract Reg from the entire sample image expThe image of the area where the sample cell is located after removing the conveyor belt is obtained, which is recorded as Img cell ; The center point of the sample cell area image is recorded as (center x , center y ), the deflection angle is recorded as phi;

[0086] Step S206, calculating a first affine transformation matrix according to the center point and deflection angle of the sample cell area image and the standard center point and standard angle 0 of the camera field of view;

[0087] In this embodiment, the first affine transformation matrix is ​​denoted as mat gstd ;

[0088] Step S207, performing affine transformation on the sample cell region image according to the first affine transformation matrix to obtain a cell region template image and save it;

[0089] In this embodiment, the template image of the cell area is denoted as Img exp ; The template image of the cell area is the cell image with the conveyor belt removed and the deflection angle of 0;

[0090] Step S208, selecting at least two positioning core templates on the battery cell area template image, and saving the position coordinates of the positioning cores;

[0091] In this embodiment, eight positioning kernel templates can be selected on the battery cell area template image, which are respectively denoted as kernel i ,1≤i≤8; the positioning core template is specifically a small rectangle, and the position coordinates of the positioning core are the upper left corner coordinates and the lower right corner coordinates of the small rectangle;

[0092] Step S30, obtaining a whole image of the image to be tested, and generating a corrected image of the cell area of ​​the image to be tested; specifically comprising:

[0093] Step S301, obtaining a whole image of a test image;

[0094] The entire image of the image to be tested can be obtained from the entire image of all cells, or can be taken again by an industrial camera;

[0095] Step S302, reading the stored binarization threshold, binarizing the entire image of the image to be tested, and obtaining the entire binarized image of the image to be tested;

[0096] Step S303, reading the stored conveyor belt position coordinates, generating a conveyor belt area according to the conveyor belt position coordinates; subtracting the conveyor belt area from the entire binary image of the image to be tested to obtain a difference area image of the image to be tested;

[0097] Step S304, reading the saved denoising parameters, and performing denoising operation on the difference area image of the image to be tested;

[0098] In this embodiment, the denoising operation is implemented by opening the operation kernel as a rectangle; the stored denoising parameters are the length and width of the operation kernel;

[0099] Step S305, performing convexity transformation on the difference area image of the image to be tested after denoising to obtain the cell area of ​​the image to be tested; intercepting the area image where the cell area of ​​the image to be tested is located from the entire image of the image to be tested to obtain the cell area image of the image to be tested after removing the conveyor belt; performing image processing on the cell area image of the image to be tested to obtain the center point and deflection angle of the cell area image of the image to be tested;

[0100] Step S306, calculating a second affine transformation matrix according to the center point and deflection angle of the cell area image of the image to be tested and the standard center point and standard angle 0 of the camera field of view;

[0101] Step S307, reading the saved positioning core template and the position coordinates of the positioning core; performing affine transformation according to the second affine transformation matrix to obtain the position coordinates of the positioning core after the affine transformation, intercepting an image near the position coordinates of the positioning core after the affine transformation on the battery cell area image of the image to be tested, and using the saved positioning core template to perform template matching on the intercepted images respectively to obtain the position coordinates of the positioning core on the battery cell area image of the image to be tested;

[0102] Step S308, calculating the transmission transformation matrix according to the position coordinates of the positioning core on the cell area image of the image to be tested and the position coordinates of the positioning core on the cell area template image of the sample;

[0103] In this embodiment, the transmission transformation matrix is ​​denoted as mat final ;

[0104] Step S309, performing transmission transformation on the cell region image of the image to be measured according to the transmission transformation matrix to obtain a cell region rectified image;

[0105] Step S40, performing a mutual difference operation on the battery cell area correction image of the image to be tested and the battery cell area template image of the sample to obtain a mutual difference operation result; specifically comprising:

[0106] Step S401, subtracting the battery cell area template image of the sample from the battery cell area rectified image of the image to be tested, and obtaining a bright defect reference image after difference;

[0107] In this embodiment, the bright defect reference image is recorded as defects. light ;

[0108] Step S402, subtracting the battery cell area correction image of the image to be tested from the battery cell area template image of the sample, and obtaining a dark defect reference image after difference;

[0109] In this embodiment, the dark defect reference image is recorded as defects. dark ;

[0110] Step S50, classifying defects according to the mutual difference operation results, and filtering defects according to the set judgment rules to determine whether the battery cell is qualified; specifically includes:

[0111] Step S501, binarize the bright defect reference image to obtain a bright defect reference binary image; traverse the connected domains therein, and for each connected domain:

[0112] If the connected domain is on the grid line, it is marked as a bright node defect;

[0113] If the connected domain is not on the grid line, it is marked as a foreign body defect;

[0114] If the connected domain is on the electrode, it is marked as an abnormal electrode defect;

[0115] Step S502, binarize the dark defect reference image to obtain a dark defect reference binary image; traverse the connected domains therein, and for each connected domain:

[0116] If the connected domain is on the gate line, it is marked as a broken gate defect;

[0117] If the connected domain is on the electrode, it is marked as an electrode missing defect;

[0118] Step S503, filtering defects according to set discrimination rules to determine whether the battery cell is qualified; in some embodiments, the set discrimination rules may include:

[0119] (1) When the gate-break length is greater than 3 pixels and the number of gate breaks is greater than 5, the cell is unqualified;

[0120] (2) If the area of ​​the foreign matter is greater than 100 pixels and the number of foreign matter is greater than 5, the cell is unqualified;

[0121] If any one of the multiple discrimination rules is satisfied, the battery cell is judged as unqualified. If all the discrimination rules are not satisfied, the battery cell is judged as qualified.

[0122] The method for detecting appearance defects of solar cells based on machine vision proposed in the above embodiments of the present invention can stably and quickly detect defects such as broken grids, bright nodes, foreign matter (dirt), and electrode abnormalities on the cells. In the face of the changing layout of the cells on the production line, the defect detection method must be able to quickly adapt to the changes in the layout and reduce the cost of layout replacement.

[0123] The embodiment of the present invention also provides a device for detecting appearance defects of solar cells based on machine vision, such as Figure 4 As shown, including:

[0124] An image acquisition module, used to acquire the entire image including the battery cell;

[0125] A first area image generation module is used to select a sample whole image without defects from all the whole images containing the battery cells, and generate a sample battery cell area template image;

[0126] The second area image generation module is used to obtain a whole image of a test image and generate a cell area correction image of the test image;

[0127] An image mutual difference module is used to perform mutual difference operation on the battery cell area correction image of the image to be tested and the battery cell area template image of the sample to obtain a mutual difference operation result;

[0128] The image defect detection module is used to classify defects according to the results of mutual difference operation, and filter the defects according to the set judgment rules to determine whether the battery cell is qualified.

[0129] like Figure 5 As shown, an embodiment of the present invention further proposes an electronic device, comprising: a processor and a memory; the processor and the memory communicate with each other, for example, are connected and communicate with each other through a bus; a computer program is stored in the memory; the processor is used to run the computer program, and when the computer program is running, the steps of the solar cell appearance defect detection method based on machine vision as described above are executed; the processor can be a CPU, or other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application-specific integrated circuits (Application Specific Integrated Circuit, ASIC), field programmable gate arrays (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components and other devices; the memory can include volatile memory, such as random access memory; the memory can also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory can also include a combination of the above-mentioned types of memory.

[0130] An embodiment of the present invention further proposes a storage medium, in which a computer program is stored, and the computer program is configured to execute the steps of the solar cell appearance defect detection method based on machine vision as described above when running; the storage medium includes a disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory (Flash Memory), a hard disk (Hard Disk Drive, abbreviated: HDD) or a solid-state drive (SSD), etc. and any combination thereof.

[0131] Finally, it should be noted that the above specific implementation methods are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

Claims

1. A method for detecting appearance defects of solar cells based on machine vision, characterized in that: The following steps are involved: Acquire the entire image including the battery cell; Select a sample whole image without defects from all the whole images containing the battery cells, and generate a template image of the battery cell area of ​​the sample; Obtain a whole image of the image to be tested, and generate a corrected image of the cell area of ​​the image to be tested; Performing a mutual difference operation on the battery cell area correction image of the image to be tested and the battery cell area template image of the sample to obtain a mutual difference operation result; Defects are classified according to the results of the mutual difference operation, and the defects are filtered according to the set judgment rules to determine whether the battery cell is qualified.

2. The method for detecting appearance defects of solar cells based on machine vision according to claim 1, characterized in that: After the entire image including the battery cell is obtained, the following steps are also included: Perform noise reduction on the entire image including the battery cell.

3. The method for detecting appearance defects of solar cells based on machine vision according to claim 1, characterized in that: The step of selecting a defect-free sample whole image from all the whole images containing the battery cells and generating a sample battery cell area template image specifically includes: Select a whole image of a sample without defects; Mark the conveyor belt area in the entire sample image and save the conveyor belt position coordinates; Binarize the entire sample image to obtain a binarized image of the entire sample; and save the binarization threshold when the image is binarized; Subtract the conveyor belt area from the entire binary image of the sample to obtain a difference area image of the sample, and perform a denoising operation on the difference area image of the sample; and record the denoising parameters in the denoising operation; The difference area image of the sample after denoising is transformed into convexity to obtain the battery cell area of ​​the sample; the area image where the battery cell area of ​​the sample is located is intercepted from the entire sample image to obtain the sample battery cell area image after removing the conveyor belt; the center point and deflection angle of the sample battery cell area image are obtained by image processing of the sample battery cell area image; Calculate a first affine transformation matrix according to the center point and deflection angle of the sample cell area image and the standard center point and standard angle 0 of the camera field of view; Performing an affine transformation on the sample cell region image according to the first affine transformation matrix to obtain a cell region template image and save the image; Select at least two positioning kernel templates on the battery cell area template image and save the position coordinates of the positioning kernels.

4. The method for detecting appearance defects of solar cells based on machine vision according to claim 3, characterized in that: The step of acquiring a whole image of a test image and generating a cell region correction image of the test image specifically includes: Get the entire image of a test image; Read the saved binarization threshold, binarize the entire image of the image to be tested, and obtain the entire binarized image of the image to be tested; Read the saved conveyor belt position coordinates, and generate the conveyor belt area according to the conveyor belt position coordinates; subtract the conveyor belt area from the entire binary image of the image to be tested to obtain the difference area image of the image to be tested; Read the saved denoising parameters and perform denoising operation on the difference area image of the image to be tested; The difference area image of the image to be tested after denoising is subjected to convex transformation to obtain the cell area of ​​the image to be tested; the area image where the cell area of ​​the image to be tested is located is intercepted from the entire image of the image to be tested to obtain the cell area image of the image to be tested after removing the conveyor belt; the center point and deflection angle of the cell area image of the image to be tested are obtained by image processing of the cell area image of the image to be tested; Calculate the second affine transformation matrix according to the center point and deflection angle of the cell area image of the image to be tested and the standard center point and standard angle 0 of the camera field of view; Read the saved positioning core template and the position coordinates of the positioning core; perform affine transformation according to the second affine transformation matrix to obtain the position coordinates of the positioning core after the affine transformation, intercept an image near the position coordinates of the positioning core after the affine transformation on the battery cell area image of the image to be tested, and use the saved positioning core template to perform template matching on the intercepted images to obtain the position coordinates of the positioning core on the battery cell area image of the image to be tested; Calculate the transmission transformation matrix according to the position coordinates of the positioning kernel on the cell region image of the image to be tested and the position coordinates of the positioning kernel on the cell region template image of the sample; The cell area image of the image to be measured is subjected to transmission transformation according to the transmission transformation matrix to obtain a cell area correction image.

5. The method for detecting appearance defects of solar cells based on machine vision according to claim 3 or 4, characterized in that: The denoising operation is implemented by opening the operation kernel as a rectangle; the denoising parameters are the length and width of the operation kernel.

6. The method for detecting appearance defects of solar cells based on machine vision according to claim 3 or 4, characterized in that: The step of performing a mutual difference operation on the cell region correction image of the image to be tested and the cell region template image of the sample to obtain a mutual difference operation result specifically includes: Subtract the template image of the cell area of ​​the sample from the corrected image of the cell area of ​​the image to be tested, and obtain the bright defect reference image after difference; The dark defect reference image is obtained by subtracting the deflection correction image of the cell area of ​​the image to be tested from the template image of the cell area of ​​the sample.

7. The method for detecting appearance defects of solar cells based on machine vision according to claim 3 or 4, characterized in that: The defect classification according to the mutual difference operation result and the filtering of the defects according to the set judgment rules to determine whether the battery cell is qualified specifically include: Binarize the bright defect reference image to obtain a bright defect reference binary image; traverse the connected domains therein, and for each connected domain: If the connected domain is on the grid line, it is marked as a bright node defect; If the connected domain is not on the grid line, it is marked as a foreign body defect; If the connected domain is on the electrode, it is marked as an abnormal electrode defect; Binarize the dark defect reference image to obtain a dark defect reference binary image; traverse the connected domains therein, and for each connected domain: If the connected domain is on the gate line, it is marked as a broken gate defect; If the connected domain is on the electrode, it is marked as an electrode missing defect; Defects are filtered according to the set judgment rules to determine whether the battery cell is qualified.

8. A solar cell appearance defect detection device based on machine vision, characterized in that: include: An image acquisition module, used to acquire the entire image including the battery cell; A first area image generation module is used to select a sample whole image without defects from all the whole images containing the battery cells, and generate a sample battery cell area template image; The second area image generation module is used to obtain a whole image of a test image and generate a cell area correction image of the test image; An image mutual difference module is used to perform mutual difference operation on the battery cell area correction image of the image to be tested and the battery cell area template image of the sample to obtain a mutual difference operation result; The image defect detection module is used to classify defects according to the results of mutual difference operation, and filter the defects according to the set judgment rules to determine whether the battery cell is qualified.

9. An electronic device, characterized in that: include: a memory storing a computer program; A processor is used to run the computer program, and when the computer program is run, the steps of the method for detecting appearance defects of solar cells based on machine vision according to any one of claims 1 to 7 are executed.

10. A storage medium, characterized in that: The storage medium stores a computer program, and the computer program is configured to execute the steps of the method for detecting appearance defects of solar cells based on machine vision according to any one of claims 1 to 7 when running.

Images