Tab dislocation detection method and apparatus, computer device, and storage medium

Abstract

The present application relates to the technical field of batteries, and provides a tab dislocation detection method and apparatus, a computer device, and a storage medium. The method comprises: carrying out key point detection on a tab image to obtain multiple key point position maps; on the basis of the multiple key point position maps, determining position information of a current tab; on the basis of the position information, determining the actual width of the current tab; and in response to a difference value between the actual width and a preset width being greater than a difference threshold, determining that the current tab is dislocated. According to the technical solution, key point detection is carried out on the tab image under front illumination, so that the coordinates of the key points of the tab are determined, and the difference value between the actual width of the tab and the preset width is determined, thereby determining whether the tab satisfies standards. Compared with the traditional methods for using a template matching algorithm to process a tab image under back illumination, the present method not only can be suitable for multiple extreme tab conditions, but also is suitable for a complex illumination environment, and has higher accuracy and robustness.

Description

Tab misalignment detection method, device, computer equipment and storage medium

[0001] This application claims priority to the Chinese patent application filed with the China Patent Office on October 17, 2023, application number 202311345364.9, and application name "Pole ear misalignment detection method, device, computer equipment and storage medium", all contents of which are incorporated by reference into this application. Technical Field

[0002] The present application relates to the field of battery technology, and in particular to a tab misalignment detection technology. Background Art

[0003] Tabs are the metal conductors that connect the positive and negative electrodes of a battery cell. They serve as the contact point between the two electrodes during charging and discharging, and are a crucial component of lithium-ion polymer batteries. During industrial manufacturing, tab misalignment can occur due to factors such as material quality, separator thickness, and cutting techniques. This can significantly impact battery performance and even lead to substandard battery quality. Therefore, detecting tab misalignment is a critical issue that needs to be addressed.

[0004] Currently, tab misalignment detection is commonly performed using a template matching algorithm. This method, based on a template matching algorithm, processes tab images captured under backlight conditions to determine the tab's corner locations and, therefore, the offset between the tab's actual width and the preset width. If the offset exceeds a threshold, the tab is considered misaligned and requires adjustment or rejection. If the offset is less than the threshold, the tab is considered acceptable. However, this method suffers from poor accuracy and robustness, and is susceptible to environmental factors such as lighting, noise, and image quality, leading to errors in the detection results.

[0005] Summary of the Invention

[0006] The present invention provides a method, apparatus, computer device, and storage medium for detecting tab misalignment, which are applicable to a variety of extreme tab conditions and complex lighting environments, and have higher accuracy and robustness. The technical solution is as follows:

[0007] In one aspect, a tab misalignment detection method is provided, the method being executed by a computer device and comprising:

[0008] Performing key point detection on the tab image to obtain a plurality of key point position maps, wherein the tab image is an image of the current tab captured based on frontal illumination, and the plurality of key point position maps are used to respectively indicate a plurality of corner points of the current tab, with one corner point corresponding to one key point position map;

[0009] Determining position information of the current tab based on the multiple key point position images, where the position information is used to indicate positions of multiple corner points of the current tab in the tab image;

[0010] Determining the actual width of the current tab based on the position information;

[0011] In response to a difference between the actual width and the preset width being greater than a difference threshold, it is determined that the current tab is misaligned.

[0012] In another aspect, a tab misalignment detection device is provided, the device being deployed on a computer device and comprising:

[0013] a detection module, configured to perform key point detection on a tab image to obtain a plurality of key point location maps, wherein the tab image is an image of the current tab captured under frontal illumination, and the plurality of key point location maps are configured to respectively indicate a plurality of corner points of the current tab, with one corner point corresponding to one key point location map;

[0014] A first determining module is configured to determine position information of the current tab based on the multiple key point position images, wherein the position information is used to indicate positions of multiple corner points of the current tab in the tab image;

[0015] A second determining module is configured to determine an actual width of the current tab based on the position information;

[0016] The third determining module is configured to determine that the current tab is misaligned in response to a difference between the actual width and the preset width being greater than a difference threshold.

[0017] On the other hand, a computer device is provided, which includes a processor and a memory, wherein the memory is used to store at least one computer program, and the at least one computer program is loaded and executed by the processor to implement the tab misalignment detection method in the embodiment of the present application.

[0018] On the other hand, a computer-readable storage medium is provided, in which at least one computer program is stored. The at least one computer program is loaded and executed by a processor to implement the tab misalignment detection method in the embodiment of the present application.

[0019] On the other hand, a computer program product is provided, including a computer program, wherein the computer program is executed by a processor to implement the tab misalignment detection method in the embodiment of the present application.

[0020] The present application provides a method for detecting tab misalignment, which performs key point detection on the tab image under frontal illumination. The tab image used in the present application is obtained by taking the tab under frontal illumination. Compared with the tab image under backal illumination, the tab image under frontal illumination can clearly distinguish the tab and the isolation film by color, which is more helpful in identifying the corner points of the tab. Therefore, the present application can more accurately detect the key point position map based on the tab image under frontal illumination, and then determine the position of the tab corner in the tab image based on the key point position map, and determine the difference between the actual width of the tab and the preset width, and then determine whether the tab is qualified. Compared with using a template matching algorithm to process the tab image under backal illumination, the present method is more helpful in identifying the corner points of the tab. It is not only applicable to a variety of extreme tab situations, but also applicable to complex lighting environments, with higher accuracy and robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0022] FIG1 is a schematic diagram of an implementation environment of a tab misalignment detection method provided according to an embodiment of the present application;

[0023] FIG2 is a flow chart of a tab misalignment detection method provided according to an embodiment of the present application;

[0024] FIG3 is a flow chart of another tab misalignment detection method provided according to an embodiment of the present application;

[0025] FIG4 is a schematic diagram of a tab corner point provided according to an embodiment of the present application;

[0026] FIG5 is a schematic diagram of an extreme tab corner point provided according to an embodiment of the present application;

[0027] FIG6 is a block diagram of a tab misalignment detection device provided according to an embodiment of the present application;

[0028] FIG7 is a schematic structural diagram of a terminal provided according to an embodiment of the present application;

[0029] FIG8 is a schematic diagram of the structure of a server provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0030] In order to make the objectives, technical solutions and advantages of this application clearer, the implementation methods of this application will be further described in detail below with reference to the accompanying drawings.

[0031] In this application, the terms "first", "second", etc. are used to distinguish identical or similar items with substantially the same effects and functions. It should be understood that there is no logical or temporal dependency between "first", "second", and "nth", nor is there any limitation on the quantity and execution order.

[0032] In the present application, the term "at least one" means one or more, and the term "plurality" means two or more.

[0033] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, storage, display, etc.), and signals involved in this application are all authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data must comply with the relevant laws, regulations, and standards of the relevant countries and regions. For example, the tab images and backlight images involved in this application were obtained with full authorization.

[0034] FIG1 is a schematic diagram of an implementation environment for a tab misalignment detection method according to an embodiment of the present application. Referring to FIG1 , the implementation environment includes a terminal 101 and a server 102. Terminal 101 and server 102 can be connected directly or indirectly via wired or wireless communication, which is not limited in this application.

[0035] In some embodiments, terminal 101 includes, but is not limited to, a mobile phone, a computer, an intelligent voice interaction device, a smart home appliance, an in-vehicle terminal, an aircraft, and the like. An application may be installed and running on terminal 101. This application can detect key points on a tab image under frontal illumination to determine whether the tab is qualified. Users can log in to this application to view the results of the tab misalignment detection. This application is associated with server 102, which provides backend services to terminal 101.

[0036] In some embodiments, server 102 is an independent physical server, or it can be a server cluster or distributed system composed of multiple physical servers. It can also be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms.

[0037] In some embodiments, the server 102 undertakes the main computing work and the terminal 101 undertakes the secondary computing work; or, the server 102 undertakes the secondary computing work and the terminal 101 undertakes the main computing work; or, the server 102 and the terminal 101 adopt a distributed computing architecture to perform collaborative computing.

[0038] FIG2 is a flow chart of a tab misalignment detection method according to an embodiment of the present application. The method is executed by a computer device. Referring to FIG2 , the method includes the following steps:

[0039] 201. Perform key point detection on the tab image to obtain multiple key point position maps. The tab image is an image of the current tab taken based on frontal illumination. The multiple key point position maps are used to respectively indicate multiple corner points of the current tab, and one corner point corresponds to one key point position map.

[0040] In an embodiment of the present application, a terminal performs key point detection on a tab image to obtain multiple key point location maps. The tab image is an image of the current tab under frontal illumination, the current tab being the tab for which tab misalignment detection is currently being performed, thereby detecting whether the tab is misaligned. Compared to an image of the current tab under backal illumination, the locations of corner points in the image of the current tab under frontal illumination are clearer. Key point detection is used to automatically detect and identify key points of specific objects in a given image. Key point detection includes key point regression and key point classification methods. In this embodiment of the present application, performing key point detection on the tab image refers to processing the tab image based on a key point regression method. Key points can be corner points, edge points, bright spots in dark areas, or dark spots in bright areas. In this embodiment of the present application, a key point can be a corner point of the current tab, that is, an edge point where the current tab contacts the battery cell body. The size of the key point location map is the same as that of the tab image. Each key point location map indicates a corner point of the current tab, with different key point location maps indicating different corner points within the current tab. The current tab may be the tab that needs to be detected for misalignment, and may be one or more tabs, which is not limited in this embodiment of the present application.

[0041] 202. Determine the position information of the current tab based on the multiple key point position images, where the position information is used to indicate the positions of multiple corner points of the current tab in the tab image.

[0042] In the embodiment of the present application, the terminal can determine the position of a corner point of the current tab in the tab image based on each key point position map. Accordingly, based on multiple key point position maps, the position of each corner point of the current tab in the tab image can be determined, thereby clarifying the position of the current tab in the tab image, that is, determining the position information of the current tab.

[0043] 203. Determine the actual width of the current tab based on the position information.

[0044] In the embodiment of the present application, the current tab can be a single tab on the battery cell or a double tab on the battery cell. In the embodiment of the present application, the current tab includes a positive tab and a negative tab, which is not limited in the embodiment of the present application. For any tab, the actual width of the tab can be determined based on the positions of the two corner points of the tab.

[0045] 204. In response to a difference between the actual width and the preset width being greater than a difference threshold, it is determined that the current tab is misaligned.

[0046] In an embodiment of the present application, when the difference between the actual width of the current pole tab and the preset width is greater than the difference threshold, the terminal determines that the current pole tab is misaligned and needs to be adjusted or scrapped. The preset width refers to the standard width of the current pole tab, and the preset width is related to the capacity of the battery to which the current pole tab belongs. For example, for batteries with a capacity of less than 2000mah (milliampere hours), the preset width of the pole tab can be 4mm (millimeter); for batteries with a capacity between 2000mah and 3000mah, the preset width of the pole tab can be 5mm; for batteries with a capacity of more than 3000mah, the preset width of the pole tab can be 6mm. The manufacturing methods of the battery cell where the current pole tab is located include stacking and winding. The current pole tabs in the battery cells manufactured by stacking and winding may be misaligned, that is, the upper structure and the lower structure do not completely overlap.

[0047] The embodiment of the present application provides a method for detecting tab misalignment, which performs key point detection on the tab image under frontal illumination. The tab image used in the present application is obtained by taking the tab image under frontal illumination. Compared with the tab image under backal illumination, the tab image under frontal illumination can clearly distinguish the tab and the isolation film by color, which is more helpful in identifying the corner points of the tab. Therefore, the present application can more accurately detect the key point position map based on the tab image under frontal illumination, and then determine the position of the tab corner in the tab image based on the key point position map, and determine the difference between the actual width of the tab and the preset width, and then determine whether the tab is qualified. Compared with using a template matching algorithm to process the tab image under backal illumination, the present method is more helpful in identifying the corner points of the tab. It is not only applicable to a variety of extreme tab situations, but also applicable to complex lighting environments, with higher accuracy and robustness.

[0048] It should be noted that the embodiments of the present application provide multiple methods for performing key point detection on a tab image to obtain multiple key point location maps. In one possible implementation, the multiple key point location maps are multiple key point heat maps. Key point detection on the tab image to obtain multiple key point location maps can be performed by processing the tab image using a key point detection model to obtain multiple key point heat maps. The key point detection model is used to perform key point detection on the input image. The key point detection model can be pre-trained.

[0049] FIG3 is a flow chart of another tab misalignment detection method provided in accordance with an embodiment of the present application. The method is executed by a computer device. Referring to FIG3 , the method includes the following steps:

[0050] 301. Obtain a training data set, where the training data set includes at least one sample image obtained by photographing a first sample tab based on frontal illumination, where the sample image is annotated with real position information of corner points corresponding to the first sample tab.

[0051] In an embodiment of the present application, a sample image can be obtained by installing a high-resolution camera on the production line to photograph the first sample tab. The corner points of the first sample tab are the edge points where the first sample tab contacts the battery cell body. Taking the first sample tab as a bipolar tab as an example, the sample image usually includes a positive tab and a negative tab. Accordingly, the first sample tab has four corner points, two for the positive tab and two for the negative tab. It should be noted that the embodiment of the present application does not limit the number and position of the corner points of the first sample tab.

[0052] The sample images in the training data set are images of the first sample tab under frontal illumination. Frontal illumination means that the illumination direction of the light source is the same as the shooting direction of the camera. Conversely, back illumination means that the illumination direction of the light source is opposite to the shooting direction of the camera. Each sample image is marked with the real position information of the corner points of the first sample tab, as shown in Figure 4, which is a schematic diagram of the tab corner points provided according to an embodiment of the present application. As shown in Figure 4, the first sample tab includes a positive tab and a negative tab, and the corner points of the first sample tab are marked as 1, 2, 3 and 4 respectively.

[0053] In some embodiments, the method of marking the real position information of the corner points of the first sample tab in the sample image can be manual marking, automatic marking, or a combination of automatic marking and manual marking. Generally, using manual marking to mark the real position information of the corner points of the first sample tab in the sample image can ensure the accuracy of the real position information of the corner points of the first sample tab. However, when the number of sample images is too large, in order to avoid high costs, the real position information of the corner points of the first sample tab in the sample image can be marked by automatic marking first, and then the wrongly marked real position information can be corrected by manual marking, thereby ensuring the accuracy of the real position information of the corner points of the first sample tab in the sample image.

[0054] In some embodiments, the training data set may include sample images indicating extreme tab corner conditions, including problems such as tab misalignment, tab curling, and tab flipping. Referring to Figure 5, Figure 5 is a schematic diagram of an extreme tab corner provided according to an embodiment of the present application. Taking the first sample tab as a single tab as an example, the image (a) in Figure 5 is an image obtained by photographing the tab based on frontal illumination, the image (b) is an image obtained by photographing the tab based on back illumination, the image (c) is a simplified schematic diagram of an image obtained by photographing the tab based on frontal illumination, and the image (d) is a simplified schematic diagram of an image obtained by photographing the tab based on back illumination. In a battery cell manufactured by winding, when the winding needle is pulled out, the problem of the isolation film being pulled out may occur, that is, the isolation film closer to the winding needle is partially pulled out along with the winding needle. As shown in image (a) of Figure 5, in an image of the tab captured using frontal illumination, the entire cell appears in a variety of grayscale values. Due to the translucent nature of the separator, the tab and separator can be distinguished by color, allowing the true location of the tab corners to be determined. For example, in image (c) of Figure 5, the left tab corner is 3, and the right tab corner is 1. As shown in image (b) of Figure 5, in an image of the tab captured using backside illumination, the entire cell appears black against a white background. Due to the lack of translucency of the separator, the tab and separator cannot be distinguished by color, making it difficult to determine the true location of the tab corners. For example, in image (d) of Figure 5, the left tab corner is 3, but it is difficult to determine whether the right tab corner is 1 or 2. This means that in some extreme cases of tab corners, images captured using backside illumination are inadequate and may make it impossible to determine the true location of the tab corners. Therefore, the sample images in the training dataset are images obtained by photographing the tabs based on frontal illumination.

[0055] It should be noted that the more sample images a training dataset contains, the better the detection results of the keypoint detection model obtained using the training dataset. The richer the variety of sample images indicating extreme corner points in the training dataset, the better the detection results of the keypoint detection model obtained using the training dataset.

[0056] In some embodiments, the sample images in the training dataset are obtained by performing data preprocessing on the captured original images. This data preprocessing includes image translation, small-angle rotation, and normalization. This data preprocessing can simulate situations such as robot arm placement offset that occur on real production lines, thereby improving the generalization capability of the key point detection model obtained using the training dataset.

[0057] 302. Process the sample image using the first to-be-trained model to obtain a plurality of sample key point heat maps, where the sample key point heat maps are used to respectively indicate a plurality of corner points of the first sample tab, and one corner point corresponds to one sample key point heat map.

[0058] During the training process, the terminal processes the sample image through the first to-be-trained model to obtain a heat map of the sample key points corresponding to each corner point.

[0059] The embodiment of the present application does not limit the model structure of the first model to be trained. In one possible implementation, the first model to be trained may be an HRnet (High-Resolution Network) model. HRnet mainly consists of two parts: a high-resolution subnetwork and a multi-scale fusion module. It should be noted that the first model to be trained may be a model that uses other key point detection algorithms, and the embodiment of the present application does not limit this.

[0060] The high-resolution subnetwork ensures that feature maps maintain high resolution throughout the model's processing, thereby avoiding issues such as information loss and blurring caused by reduced feature map resolution. The high-resolution subnetwork consists of multiple residual modules, each of which includes two convolutional layers and a skip connection. Convolutional layers extract and map features from the input data through convolution operations, outputting corresponding feature maps. Skip connections directly append the input to the output, transferring information from the input to the output. The purpose of skip connections is to avoid issues such as vanishing and exploding gradients. The input of the high-resolution subnetwork is a sample image (HxWx3), and the output is a high-resolution feature map. The sample image has dimensions HxW, meaning H for height, W for width, and 3 channels.

[0061] The multi-scale fusion module is used to fuse contextual information at different scales, that is, to fuse feature maps at multiple resolutions. It comprises sub-networks at multiple resolutions, each corresponding to a residual module. These sub-networks exchange information through upsampling and downsampling operations. The output of the multi-scale fusion module is a high-resolution feature map that incorporates information from multiple scales.

[0062] In other words, HRNet consists of multiple scale branches, each of which outputs a feature map that integrates multi-scale information. However, the final output layer of HRNet only outputs the highest-resolution feature map that integrates multi-scale information, namely the sample keypoint heat map.

[0063] For example, the input to HRNet is a sample image (64x64x3), meaning it's a three-channel sample image with a height of 64 pixels and a width of 64 pixels. After passing it through a convolutional layer, a scale branch is downsampled by a factor of 4, resulting in a feature map with a height of 32 pixels and a width of 32 pixels. Another convolutional layer results in a scale branch downsampled by a factor of 8, resulting in a feature map with a height of 16 pixels and a width of 16 pixels. Similarly, downsampled scale branches are obtained, downsampled by a factor of 16, downsampled by a factor of 32, and so on. In this process, the output of each scale branch is fused by the outputs of all branches, meaning the scale branches are fused in a fully connected fashion. At the final output, the outputs of the downsampled scale branches are upsampled by the corresponding multiples, and all the resulting outputs are summed and processed to produce the highest-resolution feature map that incorporates multi-scale information. The output of HRNet is a feature map with a height of 64 pixels and a width of 64 pixels, which is the sample keypoint heatmap.

[0064] It should be noted that in traditional tab misalignment detection methods, a model matching algorithm is used to detect the image of the tab under back lighting. When faced with some extreme tab corners, the image obtained by photographing the tab based on back lighting is insufficient, which may make it impossible to determine the true position information of the tab corner. Therefore, the sample images in the training data set are images obtained by photographing the tab based on front lighting. When using a model matching algorithm to detect tab misalignment, it is necessary to slide a standard tab template in the sample image and calculate the degree of matching to find the position that best matches the standard tab template, and then determine the position information of the tab corner. However, in the image obtained by photographing the tab based on front lighting, the surface wrinkles of the copper and aluminum tabs lead to uneven imaging brightness, that is, the entire battery cell presents multiple grayscale values ​​in the image, which interferes with the matching process. Therefore, a key point detection algorithm is used instead of a template matching algorithm.

[0065] 303. Based on the real position information annotated by the sample image and multiple sample key point heat maps, adjust the parameters of the first to-be-trained model to obtain a key point detection model.

[0066] In an embodiment of the present application, based on the annotations in the sample image, the actual position information of the extreme lug corners can be determined, and based on the sample key point heat map, the predicted position information of the extreme lug corners can be determined. The actual position information of the extreme lug corners and the predicted position information of the extreme lug corners are compared, and based on this, the parameters of the first model to be trained are adjusted to complete the training to obtain a key point detection model with better detection effect. The model structure of the key point detection model is the same as the model structure of the first model to be trained, except that the parameters are optimized.

[0067] It should be noted that in the process of training the key point detection model, the label heat map is used as the label for training. The label heat map is obtained based on the sample image. The pixel value of the key point in the label heat map is 1, and the pixel values ​​of the pixels at other positions decrease according to the Gaussian distribution. That is, the pixel value at the key point is the maximum value, and the pixel values ​​around the key point are in a ring-shaped decreasing state. It should be noted that when the pixel value of the key point in the label heat map is 1 and the pixel value of the pixel points at other positions is 0, not only will the small number of positive samples lead to an increase in the difficulty of training, but also when the true position information of the annotated extreme ear corners produces errors, it will affect the training effect of the key point detection model. Therefore, the pixel value of the key point in the label heat map is 1, and the pixel values ​​of the pixels at other positions decrease according to the Gaussian distribution.

[0068] 304. The tab image is processed by a key point detection model to obtain multiple key point heat maps. The key point detection model is used to perform key point detection on the input image.

[0069] In an embodiment of the present application, after the key point detection model is obtained through training, the terminal can process the tab image through the key point detection model to obtain the key point heat map corresponding to each corner point. Among them, the tab image is the image of the current tab under front lighting. The tab image can be obtained by installing a high-resolution camera on the production line to shoot the tab. Compared with the image of the current tab under back lighting, the position of the corner point in the image of the current tab under front lighting is clearer. Among them, the key point refers to the corner point of the current tab, that is, the edge point where the current tab contacts the battery body, which is used to indicate the width range of the current tab. The size of the key point heat map is the same as the size of the tab image, and multiple key point heat maps are used to respectively indicate multiple corner points of the current tab, each corner point corresponds to a key point heat map, and different corner points correspond to different key point heat maps.

[0070] In some embodiments, key point detection can be performed by combining features from an image of the current tab under frontal illumination and an image of the current tab under backal illumination. Accordingly, feature extraction is performed on the tab image to obtain a first feature map; feature extraction is performed on the backlit image to obtain a second feature map. The backlit image refers to an image of the current tab taken under backside illumination. The first and second feature maps are fused to obtain a fused feature map. Key point detection is performed based on the fused feature map to obtain multiple key point location maps.

[0071] Multimodal fusion of the tab image and backlight image provides more information about the current tab. The tab image and backlight image are fed into two weight-sharing feature extraction networks for feature extraction and multi-scale fusion. The resulting fused feature map is then fed into a keypoint detection model, where it is processed to produce multiple keypoint heatmaps.

[0072] In some embodiments, the tab image and the backlight image can be spliced ​​together to generate multiple key point location maps based on the resulting spliced ​​image. For example, if the backlight image is a single-channel image and the tab image is a three-channel image, splicing the backlight image and the tab image together yields a four-channel spliced ​​image. This splicing model is then input into a key point detection model, and after processing, multiple key point heat maps are generated.

[0073] 305. For any key point heat map among the multiple key point heat maps, determine a target pixel in the key point heat map, where the pixel value of the target pixel is a local maximum value in the key point heat map.

[0074] In an embodiment of the present application, for any key point heat map, the terminal determines the target pixel in the key point heat map.

[0075] In the keypoint heatmap corresponding to any corner point, the pixel value corresponding to that corner point is the highest. Therefore, we can first determine the local maximum of all pixel values ​​in the keypoint heatmap, and then determine the target pixel corresponding to this local maximum. To eliminate the influence of noise points, we use the local maximum instead of the global maximum. That is, we first denoise the keypoint heatmap and then use the peak function to determine the local maximum.

[0076] It should be noted that prediction can be performed directly to obtain the coordinates of the key points without obtaining the key point heat map and then determining the key point coordinates. Accordingly, the computational complexity and memory requirements of this method are relatively small, and the embodiments of the present application do not limit this.

[0077] The number of target pixels may be 1 or greater than 1. Accordingly, when the number of target pixels is 1, step 306 is executed; when the number of target pixels is greater than 1, steps 307 to 309 are executed.

[0078] 306. In response to the number of target pixels being 1, the coordinates of the target pixel in the key point heat map are determined as the position of a corner point of the current tab in the tab image.

[0079] In an embodiment of the present application, when the number of target pixels is 1, the terminal determines the position of the corresponding corner point of the current tab in the tab image based on the coordinates of the target pixel in the key point heat map. Among them, one target pixel corresponds to one coordinate in the key point heat map. Optionally, the origin of the coordinate axis can be located in the upper left corner of the key point heat map, with the x-axis being the horizontal direction and the y-axis being the vertical direction. Since the size of the key point heat map is the same as the size of the tab image, the coordinates of the target pixel in the key point heat map can be used as the coordinates of the corner point corresponding to the key point in the tab image, that is, the position of the corner point corresponding to the key point in the tab image can be determined.

[0080] 307. In response to the number of target pixels being greater than 1, determine the coordinates of the plurality of target pixels in the key point heat map.

[0081] In an embodiment of the present application, when the number of target pixels is greater than 1, the terminal determines the coordinates of all target pixels in the key point heat map.

[0082] 308. Interpolate the coordinates of multiple target pixels in the key point heat map to obtain sub-pixel coordinates.

[0083] In an embodiment of the present application, since the number of target pixels is greater than 1, and the coordinates of the target pixels in the key point heat map are usually integer coordinates, in order to determine the accurate coordinates and improve the prediction accuracy, the coordinates of multiple target pixels in the key point heat map can be interpolated, including bilinear interpolation and cubic spline interpolation, etc., to obtain corresponding sub-pixel coordinates, and the pixel value corresponding to the sub-pixel coordinate is the local maximum.

[0084] It should be noted that the embodiments of the present application do not limit the specific method of the interpolation operation. Among them, interpolation refers to estimating the approximate value of a function at other points by taking the value of the function at a finite number of points. Bilinear interpolation is an extension of linear interpolation on a two-dimensional rectangular grid, that is, linear interpolation is performed once in the x direction and once in the y direction. Cubic spline interpolation refers to dividing the original long sequence into several segments and constructing a cubic function for each segment so that the junctions of the segments can be smoothly connected.

[0085] 309. Determine the sub-pixel coordinates as the position of a corner point of the current tab in the tab image.

[0086] In an embodiment of the present application, since the size of the key point heat map is the same as the size of the tab image, the sub-pixel coordinates can be used as the coordinates of the corresponding corner point in the tab image, that is, the position of the corner point in the tab image can be determined.

[0087] It should be noted that when using the key point position map for key point detection, when the number of corner points at the position to be determined in the tab image changes, the key point detection model should also be retrained. That is to say, when faced with different numbers of tab corner points, different key point detection models need to be used for key point detection. In this case, in order to avoid retraining, in some embodiments, the position of the corner points of the current tab in the tab image can be determined based on the key point position map obtained by target detection. Among them, the method using target detection can not only be used for tab misalignment detection, but can also be combined with other tasks, such as folding detection; and the method using target detection is not limited by the number of corner points and is more versatile.

[0088] In some embodiments, a key point location map can be obtained through a target detection model. Accordingly, target detection is performed on the tab image through the target detection model to obtain multiple key point location maps.

[0089] In some embodiments, a target detection model can be trained first. Accordingly, a sample dataset is obtained, the sample dataset including at least one image of the second sample tab captured under frontal illumination. The images in the sample dataset are annotated with a sample annotation box, which refers to a square area centered on a corner point corresponding to the second sample tab. The images in the sample dataset are processed using a second model to be trained to obtain multiple sample key point location maps, each of which is used to indicate multiple corner points of the second sample tab, with each corner point corresponding to one sample key point location map. Based on the sample annotation boxes of the images in the sample dataset and the multiple sample key point location maps, the parameters of the second model to be trained are adjusted to obtain a target detection model. For example, after expanding 100 pixels around the corner point, the sample annotation box refers to a square area with a side length of 200 pixels centered on the corner point.

[0090] In some embodiments, the position of the corner point of the current tab in the tab image can be determined based on a labeled box in the key point location map. Accordingly, the method for determining the position information of the current tab based on multiple key point location maps can be to determine the position of the labeled box in any of the multiple key point location maps; and determine the position of the labeled box as the position of the corner point in the tab image. The center point position of the labeled box is determined as the position of the corner point in the tab image.

[0091] 310. Based on the position information, determine the actual width of the current tab.

[0092] In the embodiment of the present application, the terminal determines the actual width of the current tab based on the position information of the current tab. The current tab can be one tab on the battery cell or two tabs on the battery cell, which is not limited in the embodiment of the present application.

[0093] In some embodiments, the current tab includes a positive tab and a negative tab. The terminal determines two corner points of the positive tab and two corner points of the negative tab based on the position information; the terminal determines the Euclidean distance between the two corner points of the positive tab as the actual width of the positive tab; and the terminal determines the Euclidean distance between the two corner points of the negative tab as the actual width of the negative tab.

[0094] 311. In response to a difference between the actual width and the preset width being greater than a difference threshold, it is determined that the current tab is misaligned.

[0095] In an embodiment of the present application, when the difference between the actual width of the current tab and the preset width is greater than the difference threshold, the terminal determines that the current tab is misaligned and needs to be adjusted or scrapped. The preset width refers to the standard width of the current tab, and the difference threshold refers to the error value allowed in the tab width. For different types of tabs, the preset width and the difference threshold are not exactly the same. For example, the preset width of the tab of model A is 2mm, and the difference threshold is 0.02mm; the preset width of the tab of model B is 3mm, and the difference threshold is 0.05mm; the preset width of the tab of model C is 4mm, and the difference threshold is 0.05mm.

[0096] In some embodiments, in response to the difference between the actual width and the preset width being no greater than a difference threshold, the terminal determines that the current tab is qualified. Since errors in production are inevitable, a difference is allowed between the actual width of the current tab and the preset width. When the difference is no greater than the difference threshold, the current tab is determined to be qualified. For example, the preset width of a tab of model A is 2 mm, the difference threshold is 0.02 mm, and after using the tab misalignment detection method, the actual width of the tab of model A is determined to be 2.01 mm. At this time, the difference between the actual width of the tab of model A and the preset width is 0.01 mm, which is less than the difference threshold, indicating a reasonable error value. Therefore, the tab of model A is determined to be qualified.

[0097] It should be noted that using the key point detection model, not only can the actual width of the tab be determined, but also the relative position between the tab and the cell body can be determined. Key points are the corners of the tab and the edges of the cell body. In fact, by expanding upon the embodiments of this application, key points can be any fixed point in a standard cell, and this embodiment of this application is not limited thereto.

[0098] It should be noted that in the embodiment corresponding to Figure 3, steps 301-303 are the training process of the key point detection model, step 304 is the process of performing key point detection using the key point detection model, steps 305-306 are a method of determining the position information of the current tab based on multiple key point position maps when the multiple key point position maps are multiple key point heat maps, and steps 307-309 are another method of determining the position information of the current tab based on multiple key point position maps when the multiple key point position maps are multiple key point heat maps. Steps 301-303, step 304, steps 305-306, and steps 307-309 can each be performed separately. For example, when performing key point detection using the key point detection model, the embodiment of the present application does not limit the training method of the key point detection model. In one implementation, the key point detection model can be obtained by training through the steps shown in steps 301-303. Similarly, steps 305-306 or steps 307-309 do not depend on steps 301-304. When the multiple key point position maps are multiple key point heat maps, in one implementation method, the current tab position information can be determined through steps 305-306 or steps 307-309.

[0099] The embodiment of the present application provides a method for detecting tab misalignment, which performs key point detection on the tab image under frontal illumination. The tab image used in the present application is obtained by taking the tab image under frontal illumination. Compared with the tab image under backal illumination, the tab image under frontal illumination can clearly distinguish the tab and the isolation film by color, which is more helpful in identifying the corner points of the tab. Therefore, the present application can more accurately detect the key point position map based on the tab image under frontal illumination, and then determine the position of the tab corner in the tab image based on the key point position map, and determine the difference between the actual width of the tab and the preset width, and then determine whether the tab is qualified. Compared with using a template matching algorithm to process the tab image under backal illumination, the present method is more helpful in identifying the corner points of the tab. It is not only applicable to a variety of extreme tab situations, but also applicable to complex lighting environments, with higher accuracy and robustness.

[0100] FIG6 is a block diagram of a tab misalignment detection device according to an embodiment of the present application. The device is used to execute the steps of the above-mentioned tab misalignment detection method. Referring to FIG6 , the tab misalignment detection device includes: a detection module 601, a first determination module 602, a second determination module 603, and a third determination module 604.

[0101] Detection module 601, for performing key point detection on the tab image to obtain multiple key point location maps, where the tab image is an image of the current tab captured under frontal illumination. The multiple key point location maps are used to respectively indicate multiple corner points of the current tab, with each corner point corresponding to one key point location map;

[0102] A first determining module 602 is configured to determine position information of a current tab based on a plurality of key point position images, where the position information indicates positions of a plurality of corner points of the current tab in the tab image;

[0103] A second determining module 603 is configured to determine the actual width of the current tab based on the position information;

[0104] The third determining module 604 is configured to determine that the current tab is misaligned in response to a difference between the actual width and the preset width being greater than a difference threshold.

[0105] In some embodiments, the multiple key point position maps are multiple key point heat maps, and the detection module 601 is used to process the tab image through a key point detection model to obtain multiple key point heat maps. The key point detection model is used to perform key point detection on the input image.

[0106] In some embodiments, the detection module 601 is also used to obtain a training data set, which includes at least one sample image obtained by photographing the first sample tab based on frontal illumination, and the sample image is annotated with real position information corresponding to the corner points of the first sample tab; the sample image is processed by the first detection model to be trained to obtain multiple sample key point heat maps, and the sample key point heat maps are used to respectively indicate multiple corner points of the first sample tab, and one corner point corresponds to one sample key point heat map; based on the real position information annotated on the sample image and the multiple sample key point heat maps, the parameters of the first model to be trained are adjusted to obtain a key point detection model.

[0107] In some embodiments, the detection module 601 is also used to extract features from the tab image to obtain a first feature map; extract features from the backlight image to obtain a second feature map, where the backlight image refers to an image obtained by photographing the current tab based on back lighting; fuse the first feature map and the second feature map to obtain a fused feature map; and perform key point detection based on the fused feature map to obtain multiple key point position maps.

[0108] In some embodiments, the detection module 601 is further configured to perform target detection on the tab image using a target detection model to obtain a plurality of key point position maps.

[0109] In some embodiments, the detection module 601 is also used to obtain a sample data set, the sample data set includes at least one image obtained by photographing the second sample pole ear based on front lighting, the image in the sample data set is marked with a sample annotation box, and the sample annotation box refers to a square area centered on the corner point corresponding to the second sample pole ear; the image in the sample data set is processed by the second model to be trained to obtain multiple sample key point position maps, the sample key point position maps are used to respectively indicate multiple corner points of the second sample pole ear, and one corner point corresponds to a sample key point position map; based on the sample annotation box and the multiple sample key point position maps of the image in the sample data set, the parameters of the second model to be trained are adjusted to obtain a target detection model.

[0110] In some embodiments, the multiple key point position maps are multiple key point heat maps, and the first determination module 602 is used to determine the target pixel in the key point heat map for any key point heat map among the multiple key point heat maps, and the pixel value of the target pixel is the local maximum value in the key point heat map; in response to the number of target pixels being 1, the coordinates of the target pixel in the key point heat map are determined as the position of a corner point of the current tab in the tab image.

[0111] In some embodiments, the first determination module 602 is also used to determine the coordinates of multiple target pixels in the key point heat map in response to the number of target pixels being greater than 1; interpolate the coordinates of multiple target pixels in the key point heat map to obtain sub-pixel coordinates; and determine the sub-pixel coordinates as the position of a corner point of the current tab in the tab image.

[0112] In some embodiments, the first determination module 602 is further configured to determine the position of a marked box in any key point position map among the multiple key point position maps; and determine the position of the marked box as the position of the corner point in the tab image.

[0113] In some embodiments, the current tabs include a positive tab and a negative tab;

[0114] The second determination module 603 is used to determine the two corner points of the positive tab and the two corner points of the negative tab based on the position information; determine the Euclidean distance between the two corner points of the positive tab as the actual width of the positive tab; and determine the Euclidean distance between the two corner points of the negative tab as the actual width of the negative tab.

[0115] In some embodiments, the third determination module 604 is configured to determine that the current tab is qualified in response to a difference between the actual width and the preset width being no greater than a difference threshold.

[0116] The present application provides a tab misalignment detection device, which performs key point detection on the tab image under frontal illumination. The tab image used in the present application is obtained by taking a photo under frontal illumination. Compared with the tab image under backal illumination, the tab image under frontal illumination can clearly distinguish the tab and the isolation film by color, which is more helpful in identifying the corner points of the tab. Therefore, the present application can more accurately detect the key point position map based on the tab image under frontal illumination, and then determine the position of the tab corner in the tab image based on the key point position map, and determine the difference between the actual width of the tab and the preset width, and then determine whether the tab is qualified. Compared with using a template matching algorithm to process the tab image under backal illumination, the present device is more helpful in identifying the corner points of the tab. It is not only applicable to a variety of extreme tab situations, but also applicable to complex lighting environments, with higher accuracy and robustness.

[0117] It should be noted that the tab misalignment detection device provided in the above embodiment is only illustrated by the division of the above functional modules when running an application. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the terminal can be divided into different functional modules to complete all or part of the functions described above. In addition, the tab misalignment detection device provided in the above embodiment and the tab misalignment detection method embodiment are based on the same concept. The specific implementation process is described in the method embodiment and will not be repeated here.

[0118] FIG7 is a schematic diagram of the structure of a terminal provided according to an embodiment of the present application. Terminal 700 may be a portable mobile terminal, such as a smartphone, a tablet computer, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 player (Moving Picture Experts Group Audio Layer IV), a laptop computer, or a desktop computer. Terminal 700 may also be referred to as user equipment, a portable terminal, a laptop terminal, a desktop terminal, or other similar names.

[0119] Typically, the terminal 700 includes a processor 701 and a memory 702 .

[0120] The processor 701 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 701 may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), or PLA (Programmable Logic Array). The processor 701 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 701 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 701 may also include an AI (Artificial Intelligence) processor, which is used to process computing operations related to machine learning.

[0121] The memory 702 may include one or more computer-readable storage media, which may be non-transitory. The memory 702 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices and flash memory storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory 702 is used to store at least one computer program, which is executed by the processor 701 to implement the tab misalignment detection method provided in the method embodiment of the present application.

[0122] In some embodiments, the terminal 700 may further optionally include a peripheral device interface 703 and at least one peripheral device. Specifically, the peripheral device includes at least one of a radio frequency circuit 704 , a display screen 705 , a camera assembly 706 , an audio circuit 707 , and a power supply 708 .

[0123] In some embodiments, the terminal 700 further includes one or more sensors 709 , including but not limited to: an acceleration sensor 710 , a gyroscope sensor 711 , a pressure sensor 712 , an optical sensor 713 , and a proximity sensor 714 .

[0124] Those skilled in the art will appreciate that the structure shown in FIG. 7 does not limit the terminal 700 , and the terminal 700 may include more or fewer components than shown in the figure, or combine certain components, or adopt a different component arrangement.

[0125] FIG8 is a schematic diagram of the structure of a server provided according to an embodiment of the present application. The server 800 may vary significantly due to different configurations or performances, and may include one or more processors (Central Processing Units, CPUs) 801 and one or more memories 802, wherein the memories 802 store at least one computer program, which is loaded and executed by the processor 801 to implement the tab misalignment detection methods provided in the above-mentioned various method embodiments. Of course, the server may also have components such as a wired or wireless network interface, a keyboard, and an input / output interface for input and output. The server may also include other components for implementing device functions, which will not be described in detail here.

[0126] The present application also provides a computer-readable storage medium having at least one computer program stored therein, which is loaded and executed by a processor to implement the tab misalignment detection method of the above embodiment. For example, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), a magnetic tape, a floppy disk, or an optical data storage device.

[0127] The embodiment of the present application further provides a computer program product, including a computer program, which is executed by a processor to implement the tab misalignment detection method in the embodiment of the present application.

[0128] Those skilled in the art will understand that all or part of the steps of implementing the above embodiments may be accomplished by hardware, or may be accomplished by a program instructing the relevant hardware, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.

[0129] The above description is merely an optional embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A method for detecting tab misalignment, the method being executed by a computer device, the method comprising: Perform key point detection on the tab image to obtain a plurality of key point position maps, wherein the tab image is an image obtained by photographing the current tab based on front illumination, and the plurality of key point position maps are used to respectively indicate a plurality of corner points of the current tab, and one corner point corresponds to one key point position map; Based on the multiple key point position images, determining the position information of the current pole lug, wherein the position information is used to indicate the positions of multiple corner points of the current pole lug in the pole lug image; Based on the position information, determining the actual width of the current tab; In response to a difference between the actual width and the preset width being greater than a difference threshold, it is determined that the current tab is misaligned.

2. According to the method of claim 1, the plurality of key point position maps are a plurality of key point heat maps, and the key point detection is performed on the tab image to obtain the plurality of key point position maps, comprising: The tab image is processed by a key point detection model to obtain the multiple key point heat maps, and the key point detection model is used to perform key point detection on the input image.

3. The method according to claim 2, further comprising: Acquire a training data set, wherein the training data set includes at least one sample image obtained by photographing the first sample tab based on front illumination, wherein the sample image is annotated with real position information of a corner point corresponding to the first sample tab; Processing the sample image by a first detection model to be trained to obtain a plurality of sample key point heat maps, wherein the sample key point heat maps are used to respectively indicate a plurality of corner points of the first sample tab, and one corner point corresponds to one sample key point heat map; Based on the real position information annotated by the sample image and the multiple sample key point heat maps, the parameters of the first model to be trained are adjusted to obtain the key point detection model.

4. The method according to any one of claims 1 to 3, wherein the key point detection is performed on the tab image to obtain a plurality of key point position maps, comprising: Performing feature extraction on the tab image to obtain a first feature map; Extracting features from a backlight image to obtain a second feature map, wherein the backlight image refers to an image obtained by photographing the current tab based on backside illumination; fusing the first feature map and the second feature map to obtain a fused feature map; Key point detection is performed based on the fused feature map to obtain multiple key point position maps.

5. The method according to any one of claims 1 to 4, wherein the key point detection is performed on the tab image to obtain a plurality of key point position maps, comprising: The target detection model is used to perform target detection on the tab image to obtain the multiple key point position maps.

6. The method according to claim 5, further comprising: Acquire a sample data set, the sample data set comprising at least one image obtained by photographing the second sample pole lug based on front illumination, the image in the sample data set being marked with a sample annotation frame, the sample annotation frame being a square area centered on a corner point corresponding to the second sample pole lug; Processing the images in the sample data set by the second model to be trained to obtain a plurality of sample key point position maps, wherein the sample key point position maps are used to respectively indicate a plurality of corner points of the second sample tab, and one corner point corresponds to one sample key point position map; Based on the sample annotation boxes of the images in the sample data set and the multiple sample key point position maps, the parameters of the second model to be trained are adjusted to obtain the target detection model.

7. The method according to any one of claims 1 to 6, wherein the plurality of key point position maps are a plurality of key point heat maps, and the determining the position information of the current tab based on the plurality of key point position maps comprises: For any key point heat map among the multiple key point heat maps, determining a target pixel in the key point heat map, wherein a pixel value of the target pixel is a local maximum value in the key point heat map; In response to the number of the target pixel being 1, the coordinates of the target pixel in the key point heat map are determined as the position of a corner point of the current tab in the tab image.

8. The method according to claim 7, further comprising: In response to the number of the target pixels being greater than 1, determining coordinates of a plurality of the target pixels in the key point heat map; Interpolating the coordinates of the plurality of target pixels in the key point heat map to obtain sub-pixel coordinates; The sub-pixel coordinate is determined as the position of a corner point of the current pole lug in the pole lug image.

9. The method according to any one of claims 1 to 8, wherein determining the position information of the current tab based on the plurality of key point position maps comprises: For any key point location map among the plurality of key point location maps, determining a position of a marked box in the key point location map; The position of the marking box is determined as the position of the corner point in the tab image.

10. The method according to any one of claims 1 to 9, wherein the current electrode tabs include a positive electrode tab and a negative electrode tab; The determining the actual width of the current tab based on the position information includes: Based on the position information, determining two corner points of the positive electrode tab and two corner points of the negative electrode tab; The Euclidean distance between two corner points of the positive electrode ear is determined as the actual width of the positive electrode ear; The Euclidean distance between the two corner points of the negative electrode ear is determined as the actual width of the negative electrode ear.

11. The method according to any one of claims 1 to 10, further comprising: In response to a difference between the actual width and the preset width being not greater than a difference threshold, it is determined that the current tab is qualified.

12. A tab misalignment detection device, the device being deployed on a computer device, the device comprising: A detection module, used for performing key point detection on a tab image to obtain a plurality of key point position maps, wherein the tab image is an image obtained by photographing the current tab based on front illumination, and the plurality of key point position maps are used for respectively indicating a plurality of corner points of the current tab, and one corner point corresponds to one key point position map; A first determination module, used for determining the position information of the current pole lug based on the multiple key point position images, wherein the position information is used for indicating the positions of multiple corner points of the current pole lug in the pole lug image; A second determination module, configured to determine an actual width of the current tab based on the position information; The third determination module is used to determine that the current tab is misaligned in response to the difference between the actual width and the preset width being greater than a difference threshold.

13. A computer device, comprising a processor and a memory, wherein the memory is used to store at least one computer program, and the at least one computer program is loaded by the processor and executes the tab misalignment detection method according to any one of claims 1 to 11.

14. A computer-readable storage medium, wherein the computer-readable storage medium is used to store at least one computer program, wherein the at least one computer program is used to execute the tab misalignment detection method according to any one of claims 1 to 11.

15. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the method for detecting tab misalignment according to any one of claims 1 to 11 is implemented.

Images