Battery cell positioning method, battery cell assembly defect detection method and related device

Abstract

The present application discloses a cell positioning method, a cell assembly defect detection method and related devices, wherein the cell positioning method comprises: obtaining an original image; wherein the original image comprises an EL image of a first cell assembly, the first cell assembly comprises a plurality of first cell strings, and each first cell string comprises a plurality of first cells; performing corner point detection on a first target area of ​​the original image to obtain corner point information corresponding to at least part of the first cell string; performing rotation correction on the original image according to the corner point information corresponding to at least part of the first cell string; performing segmentation point detection on a second target area of ​​the corrected image to obtain segmentation point information corresponding to a plurality of first cells; generating positioning information of each first cell based on the corner point information corresponding to at least part of the first cell string and the segmentation point information corresponding to a plurality of first cells. By the above method, the accuracy of cell positioning can be improved.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a battery cell positioning method, a battery cell assembly defect detection method and related devices. Background Art

[0002] Solar photovoltaic, as a clean and environmentally friendly new energy, is an important development direction for renewable energy in various countries. Component EL imaging refers to imaging the cell components when they are powered on. After imaging, the defects that cannot be identified by the appearance of the components are judged manually or automatically by algorithms. After the defect judgment is completed, the location of the corresponding defective component needs to be calculated so that the repair personnel can repair the defect. The basis for determining the defect location is the positioning information of the cell strip (photovoltaic cell unit).

[0003] Most of the solutions in the prior art are to first perform distortion correction processing on the collected EL image of the battery cell assembly based on affine transformation, projection transformation or perspective principle to correct the image tilt problem; then process the battery cell image through machine vision or deep learning to determine the component defects, use gray value threshold segmentation to locate the battery cell unit, and output the battery cell unit information corresponding to the component defects.

[0004] However, the traditional grayscale threshold segmentation method has many disadvantages: the traditional threshold segmentation is subject to the influence of the overall grayscale value of the image. Some defects that cause battery failure will affect the image grayscale value, resulting in threshold segmentation errors; and the grayscale card control threshold requires manual setting of the threshold segmentation parameters, which has poor adaptability; for tilted components, correcting the component image based on projection transformation will change the overall image features and defect features, increasing the complexity of defect identification, and grayscale threshold segmentation will cause the overall offset of the positioning number, affecting the positioning accuracy. Summary of the invention

[0005] The present application provides a method, device, electronic device and storage medium for positioning a battery cell in an image, so as to improve the accuracy of positioning the battery cell.

[0006] To solve the above technical problems, the present application provides a battery cell positioning method, comprising: acquiring an original image; wherein the original image includes an EL imaging diagram of a first battery cell assembly, the first battery cell assembly includes multiple first battery strings, and each of the first battery strings includes multiple first battery cells; performing corner point detection on a first target area of ​​the original image to obtain corner point information corresponding to at least part of the first battery strings; based on the corner point information corresponding to at least part of the first battery strings, removing outliers from all corner points, and performing straight line fitting on the remaining corner points to obtain an edge straight line of the first battery cell assembly; according to the extension direction of the edge straight line of the first battery cell assembly, calculating the rotation angle corresponding to the edge straight line, and performing rotation correction on the original image according to the rotation angle to obtain a corrected image; performing segmentation point detection on a second target area of ​​the corrected image to obtain segmentation point information corresponding to the multiple first battery cells; based on the corner point information corresponding to at least part of the first battery string and the segmentation point information corresponding to the multiple first battery cells, generating positioning information for each of the first battery cells.

[0007] Among them, the first target area includes the left edge area and the right edge area of ​​the EL imaging diagram of the first battery cell assembly; the step of performing corner point detection on the first target area of ​​the original image to obtain corner point information corresponding to at least part of the first battery string includes: performing a first area cropping on the original image to obtain a first area image including the left edge area; performing corner point detection on the first area image through a corner point detection model to obtain corner point information of the left edge area; judging whether the corner point information of the left edge area meets the preset requirements; if so, taking the corner point information of the left edge area as the corner point information corresponding to at least part of the first battery string; if not, performing a second area cropping on the original image to obtain a second area image including the right edge area; performing corner point detection on the second area image through the corner point detection model to obtain corner point information of the right edge area; judging whether the corner point information of the right edge area meets the preset requirements; if so, taking the corner point information of the right edge area as the corner point information corresponding to at least part of the first battery string.

[0008] The training method of the corner detection model includes: obtaining at least one first sample image; wherein the first sample image includes an EL imaging diagram of a second battery cell assembly, and the second battery cell assembly includes a plurality of second battery strings; performing region cropping on the at least one first sample image to obtain a first sample region image including a left edge region and / or a right edge region of the EL imaging diagram of the second battery cell assembly; performing corner point marking on each demarcation point of the second battery string in the first sample region image to obtain a first sample region image with corner point labels; performing corner point detection on the first sample region image with corner point labels using the corner point detection model to obtain a corner point detection result for each of the first sample region images with corner point labels; and adjusting the network parameters of the corner point detection model using the corner point detection result and the corner point labels.

[0009] Among them, the second target area includes the upper edge area of ​​the EL imaging diagram of the first battery cell assembly; the step of performing segmentation point detection on the second target area of ​​the corrected image to obtain segmentation point information corresponding to the multiple first battery cells includes: performing second area cropping on the corrected image to obtain a third area image including the upper edge area; performing segmentation point detection on the third area image through a segmentation point detection model to obtain segmentation point information of the upper edge area, and using the segmentation point information of the upper edge area as the segmentation point information corresponding to the multiple first battery cells.

[0010] Among them, the training method of the segmentation point detection model includes: obtaining at least one second sample image; wherein the second sample image includes an EL imaging diagram of a third battery cell assembly, the third battery cell assembly includes at least one third battery string, and the third battery string includes multiple third battery cells; performing region cropping on the at least one second sample image to obtain a second sample region image containing an upper edge region of the EL imaging diagram of the third battery cell assembly; marking the segmentation points of each boundary point of the third battery cell in the second sample region image to obtain a second sample region image with segmentation point labels; performing segmentation point detection on the second sample region images with segmentation point labels using the segmentation point detection model to obtain segmentation point detection results for each second sample region image with segmentation point labels; and adjusting the network parameters of the segmentation point detection model using the segmentation point detection results and the segmentation point labels.

[0011] Among them, before the step of generating the positioning information of each first battery cell based on the corner point information corresponding to at least part of the first battery string and the segmentation point information corresponding to the multiple first battery cells, the method also includes: obtaining the width information of the first battery cell assembly, and determining the estimated distance range of the segmentation points corresponding to the first battery cell; judging whether the width distance between two adjacent segmentation points exceeds the estimated distance range of the segmentation points according to the segmentation point information corresponding to the multiple first battery cells; if so, supplementing a segmentation point between the two adjacent segmentation points according to the width distance between the two adjacent segmentation points, so as to update the segmentation point information corresponding to the multiple first battery cells.

[0012] In order to solve the above technical problems, the present application also provides a defect detection method for a battery cell assembly, comprising: obtaining a target EL imaging image of the battery cell assembly to be inspected; performing defect detection on the target EL imaging image to obtain at least one target defect; using a preset positioning method to determine the position of the target defect at a corresponding battery cell in the battery cell assembly to be inspected; wherein the preset positioning method is any of the battery cell positioning methods described above.

[0013] To solve the above technical problems, the present application also provides a battery cell positioning device, including: an acquisition module, the acquisition module is used to acquire an original image; wherein the original image includes an EL imaging diagram of a first battery cell assembly, the first battery cell assembly includes multiple first battery strings, and each of the first battery strings includes multiple first battery cells; a corner point detection module, the corner point detection module is used to perform corner point detection on a first target area of ​​the original image to obtain corner point information corresponding to at least part of the first battery strings; a correction module, the correction module is used to rotationally correct the original image according to the corner point information corresponding to at least part of the first battery strings to obtain a corrected image; a segmentation point detection module, the segmentation point detection module is used to perform segmentation point detection on a second target area of ​​the corrected image to obtain segmentation point information corresponding to the multiple first battery cells; a generation module, the generation module is used to generate positioning information of each first battery cell based on the corner point information corresponding to at least part of the first battery strings and the segmentation point information corresponding to the multiple first battery cells.

[0014] In order to solve the above technical problems, the present application also provides an electronic device, which includes: a memory and a processor coupled to each other, and the processor is used to execute program instructions stored in the memory to implement any of the above-mentioned battery cell positioning methods and / or battery cell assembly defect detection methods.

[0015] To solve the above technical problems, the present application also provides a computer-readable storage medium, which stores program instructions, and the program instructions can be executed to implement the battery cell positioning method as described in any of the above items, and / or the battery cell assembly defect detection method.

[0016] The beneficial effects of the present application are as follows: Different from the prior art, after acquiring the original image, since the original image includes the EL imaging diagram of the first battery cell assembly, and the first battery cell assembly includes multiple first battery strings, and each first battery string includes multiple first battery cells, the present application can perform corner point detection on the first target area of ​​the original image to obtain corner point information corresponding to at least part of the first battery strings; then, based on the corner point information corresponding to at least part of the first battery strings, outliers in all corner points are eliminated, and straight line fitting is performed on the remaining corner points to obtain the edge straight line of the first battery cell assembly, and according to the extension direction of the edge straight line of the first battery cell assembly, the rotation angle corresponding to the edge straight line is calculated, and the original image is rotationally corrected according to the rotation angle to obtain a corrected image; then, segmentation point detection is performed on the second target area of ​​the corrected image to obtain segmentation point information corresponding to multiple first battery cells; then, based on the corner point information corresponding to at least part of the first battery strings and the segmentation point information corresponding to multiple first battery cells, the positioning information of each first battery cell can be obtained. By adopting the detection method of the corner points between battery strings and the segmentation points between each battery cell in the battery string to locate the battery cell, the dependence of the traditional threshold segmentation method on the gray value of the image is solved, and there is no need to manually set the segmentation threshold, and the accuracy of battery cell positioning is high; in addition, the method of performing rotation correction on the original image after performing corner point straight line fitting based on the detected corner points between battery strings solves the problem of inaccurate corner point positioning under EL imaging of tilted components, thereby improving the positioning accuracy of the battery cell. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a schematic diagram of the process of the first embodiment of the battery cell positioning method provided by the present application;

[0018] Figure 2 It is a flow chart of the second embodiment of the battery cell positioning method provided by the present application;

[0019] Figure 3 yes Figure 2 A schematic diagram of a flow chart of an embodiment of step S22;

[0020] Figure 4 It is a flowchart of an embodiment of a training method for a corner detection model in the present application;

[0021] Figure 5 yes Figure 2 A schematic diagram of a flow chart of an embodiment of step S25;

[0022] Figure 6 It is a flowchart of an embodiment of a training method for a segmentation point detection model in the present application;

[0023] Figure 7a It is a schematic diagram of an EL imaging diagram of a battery cell assembly in an application scenario of the present application in a tilted state;

[0024] Figure 7b It is a schematic diagram of the EL imaging diagram of the battery cell assembly in an application scenario of the present application after being straightened;

[0025] Figure 7c It is a schematic diagram of supplementing the segmentation points of the EL imaging diagram of the battery cell assembly in an application scenario of the present application;

[0026] Figure 8 It is a flow chart of an embodiment of a defect detection method for a battery cell assembly provided by the present application;

[0027] Fig. 9 It is a structural schematic diagram of an embodiment of a battery cell positioning device provided by the present application;

[0028] Fig.10 It is a structural schematic diagram of an embodiment of an electronic device provided by the present application;

[0029] Fig.11 It is a structural schematic diagram of an embodiment of a computer-readable storage medium provided by the present application. DETAILED DESCRIPTION

[0030] The scheme of the embodiment of the present application is described in detail below in conjunction with the drawings of the specification.

[0031] In the following description, for the purpose of explanation rather than limitation, specific details such as specific system structures, interfaces, and technologies are provided to facilitate a thorough understanding of the present application.

[0032] The terms "system" and "network" are often used interchangeably in this article. The term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship. In addition, "many" in this article means two or more than two.

[0033] The cell is the basic unit of the photovoltaic cell module. It is usually made of monocrystalline silicon or polycrystalline silicon material and has the function of photoelectric conversion. The cell string is composed of multiple cells connected in series, and then multiple cell strings are connected in parallel to form a whole, which constitutes a cell module. Using the EL imaging detection method to detect the cell module, various defects of the cell module can be detected. For example, debris, cold welding, broken grid, hidden cracks, and abnormal conversion efficiency of different cells. Due to the variety of defects detected, the size of the area covered by the defects is also very different. For example, some defects may only be located in a certain cell, and some defects may cover two or more cells. Therefore, it is necessary to locate and confirm the position of the cell where the defect is located, that is, it is necessary to use a cell positioning method to locate all cells in the cell module, so as to repair the defective cells by cleaning, polishing, filling cracks and replacing cells.

[0034] See also Figure 1 , Figure 1 1 is a flow chart of the first embodiment of the battery cell positioning method provided by the present application. The battery cell positioning method in this embodiment includes the following steps:

[0035] Step S11: acquiring an original image; wherein the original image includes an EL imaging diagram of a first cell assembly, the first cell assembly includes a plurality of first cell strings, and each of the first cell strings includes a plurality of first cells.

[0036] The original image including the first cell assembly is acquired to locate the cell. In one embodiment, the user needs to detect and locate all the first cells in the first cell assembly, so the first cell assembly can be photographed by an infrared camera to obtain an EL image including the first cell assembly.

[0037] Step S12: performing corner point detection on the first target area of ​​the original image to obtain corner point information corresponding to at least part of the first battery string.

[0038] The corner points in this application are points at the corner positions in the outline of each battery string in the battery cell assembly. By detecting the corner points, the positional relationship between different battery strings can be determined. The first target area is the area where the corner points between multiple first battery strings in the first battery cell assembly are located. By performing corner point detection on the first target area of ​​the original image, the corner point information corresponding to the first battery string can be obtained, that is, the corner points between adjacent first battery strings can be quickly obtained. Compared with the original image, the network can more easily obtain the corner point features in a smaller image area. Therefore, performing corner point detection on the first target area can quickly lock the corner point information, reduce interference from other areas, reduce calculation complexity, reduce detection time, and have higher accuracy.

[0039] Step S13: based on the corner point information corresponding to at least part of the first battery string, outliers in all corner points are removed, and straight line fitting is performed on the remaining corner points to obtain the edge straight line of the first battery cell assembly.

[0040] Step S14: calculating a rotation angle corresponding to the edge straight line of the first battery cell assembly according to the extension direction of the edge straight line, and performing rotation correction on the original image according to the rotation angle to obtain a corrected image.

[0041] It can be understood that in a battery cell assembly, all battery strings are arranged side by side along the first direction, and in a single battery string, all battery cells are arranged side by side along the second direction, and the first direction is perpendicular to the second direction. Therefore, the battery cell assembly is actually an array structure composed of multiple battery cells arranged along the first direction and the second direction. Therefore, any battery cell in the battery cell assembly can be positioned and numbered along the first direction and the second direction, and each battery cell has its own positioning serial number, so that the battery cell can be positioned.

[0042] In one embodiment, when the first cell assembly is photographed by an infrared camera to obtain the original image, the first cell assembly displayed in the EL image may be tilted due to the placement angle of the first cell assembly or the shooting angle of the camera, and the EL image of the first cell assembly in a tilted state will interfere with the positioning of the cell, causing the overall positioning number of the cell to shift. Therefore, before positioning the cell, the original image needs to be tilt-corrected to correct the EL image of the first cell assembly in the original image. Since in the first cell assembly, all first battery strings are arranged side by side along the first direction, and the corner points corresponding to the first battery strings are located at the corner positions of the first battery strings in the first cell assembly, the line connecting the corner points of the first battery string is the edge straight line of the first cell assembly. It is only necessary to determine the tilt angle of the edge straight line of the first cell assembly in the original image to correct the EL image of the first cell assembly in the original image according to its tilt angle. Specifically, please refer to Figure 7a Based on the detected corner point information corresponding to the first battery string, all corner points can be fitted with straight lines. At this time, the outliers in all corner points can be removed first, and the remaining corner points can be fitted with straight lines to obtain the edge straight line of the first battery cell assembly; the extension direction of the edge straight line of the first battery cell assembly is the arrangement direction of each first battery string in the first battery cell assembly. At this time, it can be judged whether the arrangement direction of the battery string is regular in the original image and whether the first battery cell assembly is tilted according to the extension direction of the edge straight line of the first battery cell assembly. If it is tilted, the angle between the extension direction of the edge straight line of the first battery cell assembly and the standard direction (for example, the vertical direction) can be calculated. The angle is the rotation angle corresponding to the edge straight line. The original image is then rotationally corrected according to the rotation angle to obtain a corrected image. In the corrected image, the first battery cell assembly has been put in place.

[0043] Step S15: performing segmentation point detection on the second target area of ​​the corrected image to obtain segmentation point information corresponding to the plurality of first battery cells.

[0044] The segmentation point in the present application is the dividing point between adjacent cells in a cell assembly. By detecting the segmentation point, the positional relationship between different cells in a single cell string can be determined. The second target area is the area where the segmentation points between all first cells in one or more first cell strings in the first cell assembly are located. By performing segmentation point detection on the second target area of ​​the corrected image, the segmentation point information corresponding to two adjacent first cells in the first cell string can be obtained, that is, the segmentation point between adjacent first cells can be quickly obtained. Compared with the entire corrected image, the network can more easily obtain the segmentation point features in a smaller image area. Therefore, performing segmentation point detection on the second target area can quickly lock the segmentation point information, reduce interference from other areas, reduce computational complexity, reduce detection time, and have higher accuracy.

[0045] Step S16: generating positioning information of each of the first battery cells based on the corner point information corresponding to at least part of the first battery strings and the segmentation point information corresponding to the plurality of first battery cells.

[0046] It can be understood that the first direction is perpendicular to the second direction, and a plurality of first battery cells are arranged side by side along the second direction to form a first battery string, and all the first battery strings are arranged side by side along the first direction to form a first battery cell assembly. Since the first battery cell assembly has been put in place in the corrected image, the first battery cell assembly is actually an array structure of a plurality of first battery cells arranged in the horizontal and vertical directions in the corrected image. Since the corner point information corresponding to the first battery string has been obtained, the arrangement position of a first battery cell in the vertical direction can be determined through the corner points between adjacent first battery strings, and since the segmentation point information corresponding to all first battery cells in the first battery string has been obtained, the arrangement position of a first battery cell in the horizontal direction can be determined through the segmentation points between adjacent first battery cells in the first battery string, so the vertical position and horizontal position of any first battery cell in the first battery cell assembly can be determined, and the positioning information of any first battery cell can be generated. For example, any first battery cell in the first battery cell assembly can be positioned and numbered along the horizontal and vertical directions, and each first battery cell has its own positioning sequence number, so that each first battery cell can be positioned.

[0047] In the above scheme, after the original image is acquired, since the original image includes the EL imaging diagram of the first battery cell assembly, and the first battery cell assembly includes multiple first battery strings, and each first battery string includes multiple first battery cells, the first target area of ​​the original image can be detected for corner points to obtain corner point information corresponding to at least part of the first battery strings; then, based on the corner point information corresponding to at least part of the first battery strings, outliers in all corner points are eliminated, and straight line fitting is performed on the remaining corner points to obtain the edge straight line of the first battery cell assembly, and according to the extension direction of the edge straight line of the first battery cell assembly, the rotation angle corresponding to the edge straight line is calculated, and the original image is rotationally corrected according to the rotation angle to obtain a corrected image; then, segmentation point detection is performed on the second target area of ​​the corrected image to obtain segmentation point information corresponding to multiple first battery cells; then, based on the corner point information corresponding to at least part of the first battery string and the segmentation point information corresponding to multiple first battery cells, the positioning information of each first battery cell can be obtained. By adopting the detection method of the corner points between battery strings and the segmentation points between each battery cell in the battery string to locate the battery cell, the dependence of the traditional threshold segmentation method on the gray value of the image is solved, and there is no need to manually set the segmentation threshold, and the accuracy of battery cell positioning is high; in addition, according to the method of performing rotation correction on the original image after performing corner point straight line fitting based on the detected corner points between the battery strings, it is avoided that the correction of the image based on the projection transformation will change the overall characteristics of the image and the defect characteristics and increase the complexity of defect identification, and the problem of inaccurate corner point positioning under EL imaging of tilted components is solved, thereby improving the positioning accuracy of the battery cell.

[0048] See also Figure 2 , Figure 2 1 is a flow chart of a second embodiment of a cell positioning method provided in the present application. The cell positioning method in this embodiment includes the following steps:

[0049] Step S21: acquiring an original image; wherein the original image includes an EL imaging diagram of a first cell assembly, the first cell assembly includes a plurality of first cell strings, and each of the first cell strings includes a plurality of first cells.

[0050] Step S22: performing corner point detection on the first target area of ​​the original image to obtain corner point information corresponding to at least part of the first battery string.

[0051] Please combine Figure 3 , Figure 3 yes Figure 2 Schematic diagram of a flow chart of an embodiment of step S22. In one embodiment, the first target area includes the left edge area and the right edge area of ​​the EL imaging image of the first cell sheet assembly; the above step S22 specifically includes:

[0052] Step S221: performing a first region cropping on the original image to obtain a first region image including the left edge region.

[0053] Step S222: performing corner point detection on the first region image using a corner point detection model to obtain corner point information of the left edge region.

[0054] It can be understood that there is a corner point between the first battery string on the left side of the first battery cell assembly. Therefore, it is only necessary to perform a first region cropping on the left edge region of the original image containing the EL imaging diagram of the first battery cell assembly, cropping to obtain the first region image, and then perform corner point detection on the cropped first region image through the corner point detection model to obtain the corner point information of the left edge region of the first battery cell assembly. If the corner point detection model performs corner point detection on the entire original image, the complexity of the corner point detection model will be relatively high due to the large area range and much interference of the original image. In this application, the first region cropping is performed on the original image first, and the left edge region of the EL imaging diagram of the first battery cell assembly is locked as the target region, which can reduce interference, reduce the computational complexity of the corner point detection model, and improve accuracy.

[0055] Step S223: Determine whether the corner point information of the left edge area meets the preset requirements. If yes, execute step S224; if no, execute step S225.

[0056] It can be understood that the preset requirement can be that the number of detected corner points reaches a first preset value, or that the number of discrete points among the detected corner points is less than a second preset value, or that the proportion of the number of discrete points among the detected corner points to all corner points is less than a preset proportion value.

[0057] Step S224: using the corner point information of the left edge area as the corner point information corresponding to at least part of the first battery string.

[0058] Specifically, based on the corner point information of the left edge area of ​​the first battery cell assembly obtained by detection, it can be determined whether it meets the preset requirements. If so, the corner point information of the left edge area can be used as the corner point information corresponding to the first battery string in the embodiment of the present application. Subsequently, all corner points can be fitted with straight lines to obtain the edge straight line located on the left side of the first battery cell assembly, and the original image can be rotationally corrected according to the angle between the edge straight line on the left and the vertical direction, so as to obtain a corrected image in which the first battery cell assembly in the figure has been put in place.

[0059] Step S225: performing a second region cropping on the original image to obtain a second region image including the right edge region.

[0060] Step S226: performing corner point detection on the second region image using the corner point detection model to obtain corner point information of the right edge region.

[0061] It is understandable that there are also corner points between the first battery strings on the right side of the first battery cell assembly. At this time, according to the corner point information of the left edge area of ​​the first battery cell assembly obtained by detection, if it is judged that it does not meet the preset requirements, the right edge area of ​​the EL imaging diagram of the original image containing the first battery cell assembly can be further cropped to the second area, and the second area image can be cropped. Then, the corner point detection model can be used to perform corner point detection on the cropped second area image to obtain the corner point information of the right edge area of ​​the first battery cell assembly. Similarly, if the corner point detection model performs corner point detection on the entire original image, the complexity of the corner point detection model will be high due to the large area range and much interference of the original image. The present application chooses to perform the second area cropping on the original image and lock the right edge area of ​​the EL imaging diagram of the first battery cell assembly as the target area, which can reduce interference, reduce the computational complexity of the corner point detection model, and improve accuracy.

[0062] Step S227: Determine whether the corner point information of the right edge area meets the preset requirements. If it does, execute step S228. If it does not, it means that the EL image of the first cell module in the original image may have an imaging problem. At this time, if step S23 is executed to perform straight line fitting, the fitted straight line is also wrong. Therefore, fitting can be omitted directly, and the original image can be retained, or the original image can be re-acquired, and the process ends.

[0063] Step S228: using the corner point information of the right edge area as the corner point information corresponding to at least part of the first battery string.

[0064] Specifically, based on the corner point information of the right edge area of ​​the first battery cell assembly obtained by detection, it can be determined whether it meets the preset requirements. If so, the corner point information of the right edge area can be used as the corner point information corresponding to the first battery string in the embodiment of the present application. Subsequently, all corner points can be fitted with straight lines to obtain the edge straight line located on the right side of the first battery cell assembly, and the original image can be rotationally corrected according to the angle between the edge straight line on the right and the vertical direction, so as to obtain a corrected image in which the first battery cell assembly in the figure has been put in place.

[0065] It is understandable that in other embodiments, the corner point information of the right edge area of ​​the first battery cell assembly may be detected first. If it does not meet the preset requirements, the corner point information of the left edge area of ​​the first battery cell assembly may be detected. The embodiments of the present application are not limited to only detecting the corner point information of the left edge area of ​​the first battery cell assembly first, and then detecting the corner point information of the right edge area of ​​the first battery cell assembly.

[0066] See also Figure 4 , Figure 4 1 is a flow chart of an embodiment of a training method for a corner detection model in the present application. The corner detection model in the embodiment of the present application can be obtained by training using the following method. The training method for a corner detection model in the present embodiment includes the following steps:

[0067] Step S41: acquiring at least one first sample image; wherein the first sample image comprises an EL imaging diagram of a second cell sheet assembly, and the second cell sheet assembly comprises a plurality of second cell strings.

[0068] Step S42: performing region cropping on the at least one first sample image to obtain a first sample region image including a left edge region and / or a right edge region of the EL imaging image of the second cell sheet assembly.

[0069] First, the data needs to be collected and processed. For example, the second battery cell assembly can be photographed by an infrared camera to obtain an EL imaging image including the second battery cell assembly, and the image samples are screened and sorted to obtain a first sample image; then the first sample image is cropped, and the left edge area and / or the right edge area of ​​the EL imaging image of the second battery cell assembly is locked as the target area to obtain a first sample area image as a sample set for training the corner detection model.

[0070] Step S43: marking corner points of each demarcation point of the second battery string in the first sample region image to obtain a first sample region image with corner point labels.

[0071] For example, for each dividing point of the second battery string, the corner points are directly marked on the first sample area image to obtain the first sample area image with corner point labels, so as to convert the image data into the YOLO data set format containing label information as a training set and a test set that can be used for neural network training.

[0072] Step S44: using the corner point detection model to perform corner point detection on the first sample region images with corner point labels, to obtain a corner point detection result for each of the first sample region images with corner point labels.

[0073] Step S45: using the corner point detection result and the corner point label, adjusting the network parameters of the corner point detection model.

[0074] Specifically, an initial corner detection model is constructed, and the first sample area image with corner point labels is divided into a training set and a test set according to a certain ratio. The training set is input into the initial corner detection model, and the corner detection result corresponding to the training set is output. According to the difference between the corner detection result and the corresponding corner point label, the network parameters of the corner detection model are adjusted to update the corner detection model, and the updated corner detection model is tested and verified using the test set. It can be understood that during training, the initial corner detection model outputs a predicted frame based on the initial anchor frame, and then compares it with the real frame (corner point label), calculates the difference between the two, and then reversely updates and iterates the network parameters of the model. In addition, during each training, the optimal anchor frame value corresponding to different training sets can be adaptively calculated to complete the training optimization of the corner detection model.

[0075] In an embodiment of the present application, a corner point detection model is designed to detect the EL imaging image of a battery cell assembly, and the left and right edge areas of the first sample image are cropped, and the corner points of the cropped first sample area image are marked. The corner point marking target is the dividing point between two adjacent battery strings in the battery cell assembly. After the marking is completed, the corner point labels are matched with the cropped first sample area image to generate a training set for training the corner point detection model. The resulting corner point detection model has low computational complexity, less detection time, and high accuracy.

[0076] Step S23: based on the corner point information corresponding to at least part of the first battery string, outliers in all corner points are removed, and straight line fitting is performed on the remaining corner points to obtain the edge straight line of the first battery cell assembly.

[0077] Step S24: calculating a rotation angle corresponding to the edge straight line of the first battery cell assembly according to the extension direction of the edge straight line, and performing rotation correction on the original image according to the rotation angle to obtain a corrected image.

[0078] Step S25: performing segmentation point detection on the second target area of ​​the corrected image to obtain segmentation point information corresponding to the plurality of first battery cells.

[0079] Please combine Figure 5 , Figure 5 yes Figure 2 Schematic diagram of a flow chart of an embodiment of step S25. In one embodiment, the second target area includes the upper edge area of ​​the EL imaging image of the first cell sheet assembly; the above step S25 specifically includes:

[0080] Step S251: performing a second region cropping on the corrected image to obtain a third region image including the upper edge region.

[0081] Step S252: performing segmentation point detection on the third region image by using a segmentation point detection model to obtain segmentation point information of the upper edge region, and using the segmentation point information of the upper edge region as segmentation point information corresponding to the plurality of first battery cells.

[0082] Understandably, please combine Figure 7bIn the corrected image, the first cell assembly has been put in place, so there is a segmentation point between the first cells on the upper side of the first cell assembly. Therefore, it is only necessary to perform a third area cropping on the upper edge area of ​​the EL imaging diagram of the first cell assembly in the corrected image, crop the third area image, and then perform segmentation point detection on the cropped third area image through the segmentation point detection model to obtain the segmentation point information of the upper edge area of ​​the first cell assembly. If the segmentation point detection model performs segmentation point detection on the entire corrected image, the complexity of the segmentation point detection model will be relatively high due to the large area range of the corrected image and much interference. The present application first performs a third area cropping on the corrected image, and locks the upper edge area of ​​the EL imaging diagram of the first cell assembly as the target area, which can reduce interference, reduce the computational complexity of the segmentation point detection model, improve detection speed, and have higher accuracy.

[0083] It can be understood that the embodiments of the present application are not limited to cropping only the upper edge area of ​​the corrected image. In other embodiments, the second target area can also be the lower edge area of ​​the EL imaging image including the first battery cell assembly. The regional image is obtained by cropping the lower edge area of ​​the EL imaging image including the first battery cell assembly in the corrected image. The segmentation point information of the lower edge area of ​​the first battery cell assembly can also be obtained by performing segmentation point detection through a segmentation point detection model.

[0084] See also Figure 6 , Figure 6 1 is a flow chart of an embodiment of a training method for a segmentation point detection model in the present application. The segmentation point detection model in the embodiment of the present application can be obtained by training using the following method. The training method for the segmentation point detection model in the present embodiment includes the following steps:

[0085] Step S61: acquiring at least one second sample image; wherein the second sample image comprises an EL imaging diagram of a third cell assembly, the third cell assembly comprises at least one third cell string, and the third cell string comprises a plurality of third cells.

[0086] Step S62: performing region cropping on the at least one second sample image to obtain a second sample region image including an upper edge region of the EL imaging image of the third cell sheet assembly.

[0087] First, the data needs to be collected and processed. For example, the third battery cell assembly can be photographed by an infrared camera to obtain an EL imaging image including the third battery cell assembly, and the image samples are screened and sorted to obtain a second sample image; then, the second sample image is region cropped, and the upper edge area of ​​the EL imaging image of the third battery cell assembly is locked as the target area to obtain a second sample region image as a sample set for training the segmentation point detection model.

[0088] Step S63: marking the dividing point of each of the third battery cells in the second sample region image as a segmentation point, to obtain a second sample region image with segmentation point labels.

[0089] For example, for the dividing point of each third battery cell in the third battery string, the segmentation point is directly marked on the second sample area image to obtain the second sample area image with the segmentation point label, so as to convert the image data into the YOLO data set format containing label information as a training set and a test set that can be used for neural network training.

[0090] Step S64: using the segmentation point detection model to perform segmentation point detection on the second sample region images with segmentation point labels, and obtaining segmentation point detection results for each of the second sample region images with segmentation point labels.

[0091] Step S65: using the segmentation point detection result and the segmentation point label, adjusting the network parameters of the segmentation point detection model.

[0092] Specifically, an initial segmentation point detection model is constructed, and the second sample area image with segmentation point labels is divided into a training set and a test set according to a certain ratio. The training set is input into the initial segmentation point detection model, and the segmentation point detection result corresponding to the training set is output. According to the difference between the segmentation point detection result and the corresponding segmentation point label, the network parameters of the segmentation point detection model are adjusted to update the segmentation point detection model, and the updated segmentation point detection model is tested and verified using the test set. It can be understood that during training, the initial segmentation point detection model outputs a predicted frame based on the initial anchor frame, and then compares it with the real frame (segmentation point label), calculates the difference between the two, and then reversely updates and iterates the network parameters of the model. In addition, during each training, the optimal anchor frame value corresponding to different training sets can be adaptively calculated to complete the training optimization of the segmentation point detection model.

[0093] In an embodiment of the present application, a segmentation point detection model is designed to detect the EL imaging diagram of the battery cell assembly, and the upper edge area of ​​the second sample image is cropped, and the segmentation points are marked on the cropped second sample area image. The segmentation point marking target is the dividing point between two adjacent battery cells in a single battery string in the battery cell assembly. After the marking is completed, the segmentation point label is matched with the cropped second sample area image to generate a training set for training the segmentation point detection model. The resulting segmentation point detection model has low computational complexity, less detection time, and high accuracy.

[0094] Steps S21 to S25 in this embodiment are substantially the same as steps S11 to S15 in the above embodiment, and are not described in detail herein.

[0095] Step S26: Obtain the width information of the first battery cell assembly, and determine the estimated distance range of the segmentation points corresponding to the first battery cell.

[0096] It can be understood that for the first battery cell assembly that needs to be defect detected, its size is known, and the number of first battery cells in a single first battery string that constitutes the first battery cell assembly is also known. Therefore, based on the known width information of the first battery cell assembly and the number of first battery cells arranged along the width direction of the first battery cell assembly (i.e., the second direction mentioned above), there is a first battery cell between two adjacent dividing points, and the distance size between the two adjacent dividing points can be estimated. Based on factors such as the size error of the first battery cell, the connection spacing between the first battery cells, and the installation error, the estimated distance size can be a range value, that is, the estimated distance range of the dividing points corresponding to the first battery cell is obtained.

[0097] Step S27: According to the segmentation point information corresponding to the plurality of first battery cells, determine whether the width distance between two adjacent segmentation points exceeds the segmentation point estimated distance range. If the width distance between two adjacent segmentation points exceeds the segmentation point estimated distance range, execute step S28; if the width distance between two adjacent segmentation points does not exceed the segmentation point estimated distance range, it means that there is no defect occlusion problem between the two adjacent segmentation points detected, and the segmentation point supplementation strategy of step S28 is not required at this time, and step S29 is directly executed.

[0098] Step S28: adding a division point between the two adjacent division points according to the width distance between the two adjacent division points, so as to update the division point information corresponding to the plurality of first battery cells.

[0099] Please combine Figure 7cIt can be understood that after obtaining the estimated distance range of the segmentation points corresponding to the first battery cell, the theoretical interval distance range between the two adjacent segmentation points is determined, so the actual width distance between the two adjacent segmentation points can be calculated. For example, the width distance between the two adjacent segmentation points can be calculated in turn according to the segmentation point detection frame, and then it is determined whether the width distance between the two adjacent segmentation points exceeds the estimated distance range of the segmentation points; if the actual width distance between the two adjacent segmentation points exceeds the estimated distance range of the segmentation points, it means that there is an omission in the detection between the two segmentation points, for example, there is a segmentation point in the middle that is blocked by a defect and the segmentation point is not identified. At this time, the segmentation point can be supplemented between the two segmentation points according to the difference between the actual width distance and the estimated distance range of the segmentation point, which can improve the correctness of the segmentation point detection. After the segmentation points are supplemented and the serial number is updated, the accuracy of the subsequent battery cell positioning can be guaranteed.

[0100] Step S29: generating positioning information of each of the first battery cells based on the corner point information corresponding to at least part of the first battery strings and the segmentation point information corresponding to the plurality of first battery cells.

[0101] Step S29 in this embodiment is substantially the same as step S16 in the above embodiment and will not be described in detail here.

[0102] The battery cell positioning method of the embodiment of the present application takes into account the possibility of omissions during segmentation point detection. If a segmentation point is missed due to defect occlusion or non-recognition, it will cause errors in the final battery cell positioning. Therefore, the present application can perform segmentation point detection on the second target area of ​​the corrected image, and after quickly acquiring the segmentation point corresponding to the first battery cell, determine whether there is an omission between two adjacent segmentation points based on the difference between the actual width distance between the two adjacent segmentation points and the estimated distance range of the segmentation points, and supplement the segmentation points when there are omissions. After supplementing the segmentation points, the segmentation point information corresponding to the detected first battery cell is updated to ensure the accuracy of subsequent battery cell positioning.

[0103] See also Figure 8 , Figure 8 1 is a flow chart of an embodiment of a defect detection method for a battery cell assembly provided by the present application. The defect detection method for a battery cell assembly in this embodiment includes the following steps:

[0104] Step S81: Obtain a target EL image of the battery cell assembly to be inspected.

[0105] Step S82: performing defect detection on the target EL image to obtain at least one target defect.

[0106] Step S83: Determine the position of the target defect at the corresponding cell in the cell assembly to be inspected by using a preset positioning method, wherein the preset positioning method is the cell positioning method in any of the above embodiments.

[0107] By using the EL detection method to detect defects in the battery cell assembly, the target defect can be detected through the target EL imaging image of the battery cell assembly to be detected, and then the position of the battery cell where the target defect is located can be determined and marked by the battery cell positioning method in any of the above embodiments to facilitate the subsequent process of repairing the battery cell assembly.

[0108] In the defect detection method for the battery cell assembly of the present application, the battery cells are positioned by adopting a detection method of the corner points between battery strings and the segmentation points between each battery cell in the battery string, thereby solving the dependence of the traditional threshold segmentation method on the grayscale value of the image, and there is no need to manually set the segmentation threshold, and the accuracy of battery cell positioning is high; in addition, the method of performing rotation correction on the original image after performing corner point straight line fitting based on the detected corner points between the battery strings solves the problem of inaccurate corner point positioning under EL imaging of tilted components, thereby improving the positioning accuracy of the battery cells.

[0109] See also Fig. 9 , Fig. 9 : is a schematic diagram of the structure of an embodiment of a cell positioning device provided by the present application. The cell positioning device 90 in this embodiment includes an acquisition module 900, a corner point detection module 902, a correction module 904, a segmentation point detection module 906 and a generation module 908 that are interconnected. The acquisition module 900 is used to acquire an original image; wherein the original image includes an EL imaging diagram of a first cell assembly, the first cell assembly includes a plurality of first battery strings, and each of the first battery strings includes a plurality of first cells. The corner point detection module 902 is used to perform corner point detection on a first target area of ​​the original image to obtain corner point information corresponding to at least part of the first battery string. The correction module 904 is used to perform rotation correction on the original image according to the corner point information corresponding to at least part of the first battery string to obtain a corrected image. The segmentation point detection module 906 is used to perform segmentation point detection on a second target area of ​​the corrected image to obtain segmentation point information corresponding to the plurality of first cells. The generating module 908 is configured to generate positioning information of each of the first battery cells based on the corner point information corresponding to at least part of the first battery strings and the segmentation point information corresponding to the plurality of first battery cells.

[0110] In one embodiment, the first target area includes a left edge area and a right edge area of ​​the EL imaging diagram of the first battery cell assembly; the corner detection module 902 performs a step of performing corner detection on the first target area of ​​the original image to obtain corner information corresponding to at least part of the first battery string, specifically including: performing a first area cropping on the original image to obtain a first area image including the left edge area; performing corner detection on the first area image using a corner detection model to obtain corner information of the left edge area; judging whether the corner information of the left edge area meets a preset requirement; if so, using the corner information of the left edge area as the corner information corresponding to at least part of the first battery string; if not, performing a second area cropping on the original image to obtain a second area image including the right edge area; performing corner detection on the second area image using the corner detection model to obtain corner information of the right edge area; judging whether the corner information of the right edge area meets the preset requirement; if so, using the corner information of the right edge area as the corner information corresponding to at least part of the first battery string.

[0111] In one embodiment, the second target area includes the upper edge area of ​​the EL imaging diagram of the first battery cell assembly; the segmentation point detection module 906 performs segmentation point detection on the second target area of ​​the corrected image to obtain the segmentation point information corresponding to the multiple first battery cells, specifically including: cropping the second area of ​​the corrected image to obtain a third area image including the upper edge area; performing segmentation point detection on the third area image through a segmentation point detection model to obtain the segmentation point information of the upper edge area, and using the segmentation point information of the upper edge area as the segmentation point information corresponding to the multiple first battery cells.

[0112] In one embodiment, before the generation module 908 executes the step of generating the positioning information of each first battery cell based on the corner point information corresponding to at least part of the first battery string and the segmentation point information corresponding to the plurality of first battery cells, the segmentation point detection module 906 is further used to obtain the width information of the first battery cell assembly and determine the estimated distance range of the segmentation points corresponding to the first battery cell; based on the segmentation point information corresponding to the plurality of first battery cells, determine whether the width distance between two adjacent segmentation points exceeds the estimated distance range of the segmentation points; if so, add a segmentation point between the two adjacent segmentation points based on the width distance between the two adjacent segmentation points to update the segmentation point information corresponding to the plurality of first battery cells.

[0113] See also Fig.10 , Fig.10It is a schematic diagram of the structure of an embodiment of an electronic device of the present application. The electronic device 100 in this embodiment includes a processor 1002 and a memory 1001 connected to each other; the memory 1001 is used to store program instructions, and the processor 1002 is used to execute the program instructions stored in the memory 1001 to implement the steps of any of the above-mentioned battery cell positioning method embodiments, and / or the steps of the above-mentioned battery cell assembly defect detection method embodiments. In a specific implementation scenario, the electronic device 100 may include but is not limited to: a microcomputer, a server.

[0114] Specifically, the processor 1002 is used to control itself and the memory 1001 to implement the steps of any of the above-mentioned battery cell positioning method embodiments, and / or the steps of the above-mentioned battery cell assembly defect detection method embodiments. The processor 1002 can also be called a CPU (Central Processing Unit). The processor 1002 may be an integrated circuit chip with signal processing capabilities. The processor 1002 can also be a general-purpose processor, a digital signal processor (Digital Signal Processor, DSP), an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field-programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. In addition, the processor 1002 can be implemented by an integrated circuit chip.

[0115] See also Fig.11 , Fig.11 The computer-readable storage medium 110 of the present application stores program instructions 1100 thereon, which, when executed by a processor, implement the steps of any of the above-mentioned battery cell positioning method embodiments and / or the steps of the above-mentioned battery cell assembly defect detection method embodiments.

[0116] The computer-readable storage medium 110 may specifically be a medium that can store program instructions 1100, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or the like, or may be a server storing the program instructions 1100. The server may send the stored program instructions 900 to other devices for execution, or may also execute the stored program instructions 1100 by itself.

[0117] In the several embodiments provided in the present application, it should be understood that the disclosed methods, devices and apparatuses can be implemented in other ways. For example, the above-described device and apparatus implementation methods are merely schematic. For example, the division of modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0118] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0119] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0120] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of each implementation method of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program code.

[0121] If the technical solution of this application involves personal information, the product using the technical solution of this application has clearly informed the personal information processing rules and obtained the individual's voluntary consent before processing the personal information. If the technical solution of this application involves sensitive personal information, the product using the technical solution of this application has obtained the individual's separate consent before processing the sensitive personal information, and at the same time meets the "explicit consent" requirement. For example, on personal information collection devices such as cameras, clear and prominent signs are set to inform that the personal information collection scope has been entered and personal information will be collected. If the individual voluntarily enters the collection scope, it is deemed that he or she agrees to the collection of his or her personal information; or on the device that processes personal information, the personal information processing rules are notified by obvious signs / information, and the individual's authorization is obtained through pop-up information or by asking the individual to upload his or her personal information; among them, the personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the type of personal information processed.

Claims

1. A battery cell positioning method, characterized in that: include: Acquire an original image; wherein the original image includes an EL image of a first cell assembly, the first cell assembly includes a plurality of first cell strings, and each of the first cell strings includes a plurality of first cells; Performing corner point detection on a first target area of ​​the original image to obtain corner point information corresponding to at least part of the first battery string; Based on the corner point information corresponding to at least part of the first battery string, outliers in all corner points are eliminated, and straight line fitting is performed on the remaining corner points to obtain an edge straight line of the first battery cell assembly; According to the extension direction of the edge straight line of the first battery cell assembly, a rotation angle corresponding to the edge straight line is calculated, and the original image is rotationally corrected according to the rotation angle to obtain a corrected image; Performing segmentation point detection on the second target area of ​​the corrected image to obtain segmentation point information corresponding to the plurality of first battery cells; Based on the corner point information corresponding to at least part of the first battery strings and the segmentation point information corresponding to the plurality of first battery cells, the positioning information of each of the first battery cells is generated.

2. The battery cell positioning method according to claim 1, characterized in that: The first target area includes a left edge area and a right edge area of ​​the EL imaging image of the first cell sheet assembly; The step of performing corner point detection on the first target area of ​​the original image to obtain corner point information corresponding to at least part of the first battery string includes: Performing a first region cropping on the original image to obtain a first region image including the left edge region; Performing corner point detection on the first region image by using a corner point detection model to obtain corner point information of the left edge region; Determine whether the corner point information of the left edge area meets the preset requirements; If the conditions are met, the corner point information of the left edge area is used as the corner point information corresponding to at least part of the first battery string; If not, performing a second region cropping on the original image to obtain a second region image including the right edge region; Performing corner point detection on the second region image by using the corner point detection model to obtain corner point information of the right edge region; Determine whether the corner point information of the right edge area meets the preset requirement; If the conditions are met, the corner point information of the right edge area is used as the corner point information corresponding to at least part of the first battery string.

3. The battery cell positioning method according to claim 2, characterized in that: The training method of the corner detection model includes: Acquire at least one first sample image; wherein the first sample image includes an EL imaging diagram of a second battery cell assembly, and the second battery cell assembly includes a plurality of second battery strings; Performing regional cropping on the at least one first sample image to obtain a first sample region image including a left edge region and / or a right edge region of the EL imaging image of the second cell sheet assembly; Marking corner points of each demarcation point of the second battery string in the first sample area image to obtain a first sample area image with corner point labels; Using the corner point detection model to perform corner point detection on the first sample area images with corner point labels, to obtain a corner point detection result for each of the first sample area images with corner point labels; The network parameters of the corner point detection model are adjusted using the corner point detection result and the corner point label.

4. The battery cell positioning method according to claim 1, characterized in that: The second target area includes an upper edge area of ​​the EL imaging image of the first cell sheet assembly; The step of performing segmentation point detection on the second target area of ​​the corrected image to obtain segmentation point information corresponding to the plurality of first battery cells includes: Performing a second region cropping on the corrected image to obtain a third region image including the upper edge region; The segmentation point detection is performed on the third region image by using a segmentation point detection model to obtain segmentation point information of the upper edge region, and the segmentation point information of the upper edge region is used as segmentation point information corresponding to the plurality of first battery cells.

5. The battery cell positioning method according to claim 4, characterized in that: The training method of the segmentation point detection model includes: Acquire at least one second sample image; wherein the second sample image includes an EL imaging diagram of a third cell sheet assembly, the third cell sheet assembly includes at least one third cell string, and the third cell string includes a plurality of third cells; Performing regional cropping on the at least one second sample image to obtain a second sample regional image including an upper edge region of the EL imaging image of the third cell sheet assembly; Marking the dividing point of each of the third battery cells in the second sample area image as a segmentation point to obtain a second sample area image with segmentation point labels; Using the segmentation point detection model to perform segmentation point detection on the second sample region images with segmentation point labels, to obtain segmentation point detection results for each of the second sample region images with segmentation point labels; The network parameters of the segmentation point detection model are adjusted using the segmentation point detection result and the segmentation point label.

6. The battery cell positioning method according to claim 1, characterized in that: Before the step of generating the positioning information of each of the first battery cells based on the corner point information corresponding to at least part of the first battery strings and the segmentation point information corresponding to the plurality of first battery cells, the method further includes: Acquire width information of the first battery cell assembly, and determine an estimated distance range of the segmentation points corresponding to the first battery cell; According to the segmentation point information corresponding to the plurality of first battery cells, determining whether a width distance between two adjacent segmentation points exceeds an estimated distance range of the segmentation points; If so, a division point is added between the two adjacent division points according to the width distance between the two adjacent division points, so as to update the division point information corresponding to the plurality of first battery cells.

7. A defect detection method for a battery cell assembly, characterized in that: include: Obtain a target EL image of the battery cell assembly to be inspected; Performing defect detection on the target EL image to obtain at least one target defect; Determine the position of the target defect at the corresponding battery cell in the battery cell assembly to be inspected by using a preset positioning method; Wherein, the preset positioning method is the battery cell positioning method according to any one of claims 1 to 6.

8. A battery cell positioning device, characterized in that: include: An acquisition module, the acquisition module is used to acquire an original image; wherein the original image includes an EL image of a first cell assembly, the first cell assembly includes a plurality of first cell strings, and each of the first cell strings includes a plurality of first cells; A corner point detection module, the corner point detection module is used to perform corner point detection on a first target area of ​​the original image to obtain corner point information corresponding to at least part of the first battery string; A correction module, the correction module being used to perform rotation correction on the original image according to the corner point information corresponding to at least part of the first battery string to obtain a corrected image; A segmentation point detection module, the segmentation point detection module is used to perform segmentation point detection on the second target area of ​​the corrected image to obtain segmentation point information corresponding to the plurality of first battery cells; A generating module, the generating module is used to generate positioning information of each of the first battery cells based on the corner point information corresponding to at least part of the first battery strings and the segmentation point information corresponding to the plurality of first battery cells.

9. An electronic device, characterized in that: The electronic device comprises: a memory and a processor coupled to each other, the processor being used to execute program instructions stored in the memory to implement the battery cell positioning method as described in any one of claims 1 to 6, and / or the battery cell assembly defect detection method as described in claim 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program instructions, and the program instructions can be executed to implement the battery cell positioning method according to any one of claims 1 to 6, and / or the battery cell assembly defect detection method according to claim 7.

Images