A method, device, equipment and medium for tab anomaly detection in a battery cell production process

Abstract

The application discloses a kind of for the method, device, equipment and medium of tab abnormality detection in battery cell production process, comprising: the tab area image collected in battery cell production process is segmented to generate tab mask;Based on the tab mask extraction multi-scale pixel-level feature vector, and the multi-scale pixel-level feature vector is input into unsupervised probability GMM model to identify pixel-level abnormal point;The pixel-level abnormal point is analyzed to obtain candidate abnormal area, and the parameter index of the candidate abnormal area is obtained;Determine whether the parameter index meets the set tab abnormality judgment rule, if meet, it is judged as tab abnormality, and tab abnormality result is output.Therefore, the application can be widely applied to the detection of tab folding, crease, fracture, misplacement and other abnormalities on battery cell production line.

Description

Technical Field

[0001] This invention relates to the fields of battery manufacturing and machine vision inspection, specifically to a method, apparatus, equipment, and medium for detecting abnormal tabs during the cell production process. Background Technology

[0002] The tabs are critical components connecting the internal and external electrical connections of lithium batteries. Their welding, bending, and bonding directly affect the battery's electrical performance, mechanical reliability, and safety. During winding, stacking, and modular assembly processes, tabs are prone to problems such as folding (wrinkles, warping, air bubbles, or folding outside the specified direction), misalignment, tearing, or localized damage. Traditional manual inspection is time-consuming and highly subjective. Traditional rule-based visual methods (thresholding, edge detection, template matching) are greatly affected by lighting, reflection, and tab material (copper / aluminum foil), resulting in high rates of missed and false detections.

[0003] In recent years, deep learning object detection (such as the YOLO series) and semantic segmentation have made significant progress. However, for defects such as electrode folds, which are small, localized, and visually very similar to normal areas in color and texture, deep learning models rely on a large number of labeled samples and are often unable to reliably identify them due to downsampling and receptive field limitations on small structures. In addition, electrode aberration samples are scarce in production lines, making the promotion of fully supervised methods costly and difficult to meet the requirements of real-time online operation and adaptation to limited samples.

[0004] In summary, existing electrode anomaly detection technologies mainly suffer from the following problems: 1) Difficulty in microstructure recognition: Traditional target detection networks lose minute details during scale downsampling, making it difficult to identify folded edges or shallow wrinkles with widths of only a few pixels. 2) High annotation costs: Anomaly samples are scarce, resulting in high annotation costs, and fully supervised methods struggle to obtain robust models. 3) Sensitivity to illumination and reflection: Electrode materials are metallic and highly reflective, with small differences in a single color space, making them susceptible to detection thresholds. 4) Real-time performance and industrial deployment: Complex network inference is time-consuming, making it difficult to meet production line cycle time requirements on megapixel-level images. 5) Incomplete utilization of multimodal information: When using only RGB images, the micro-geometric morphological information (height / depth, normal) of electrode folds is lost, leading to misjudgments. Summary of the Invention

[0005] The present invention aims to at least solve one of the technical problems existing in the prior art. Therefore, in response to the above-mentioned problems, the object of the present invention is to provide a method, apparatus, equipment, and medium for detecting tab anomalies during the battery cell production process, capable of achieving high-precision detection of tab anomalies.

[0006] To achieve the above-mentioned objectives, the technical solution adopted by this invention is as follows:

[0007] In a first aspect, the present invention provides a method for detecting tab abnormalities during battery cell production, comprising: The image of the tab area collected during the battery cell production process is segmented to generate a tab mask; Based on the aforementioned tab mask, multi-scale pixel-level feature vectors are extracted, and the multi-scale pixel-level feature vectors are input into an unsupervised probabilistic GMM model to identify pixel-level outliers. The pixel-level anomalies are analyzed to obtain candidate anomaly regions, and the parameter indices of the candidate anomaly regions are obtained. Determine whether the parameter index meets the set electrode abnormality judgment rules. If it does, the electrode is judged to be abnormal and the electrode abnormality result is output.

[0008] Some possible implementations also include a step of acquiring images of the tab area during the cell manufacturing process, the process of which is as follows: Use an industrial camera to collect data on the electrode area; Collect point cloud or depth images of the electrodes.

[0009] In some possible implementations, the image of the tab area acquired during the battery cell manufacturing process is segmented to generate a tab mask, including: A lightweight segmentation network is used to locate the large bounding box of the pole ears in the image of the pole ear region to obtain a sub-image; The sub-image is segmented using a pre-trained SAM model to obtain the pole ear mask.

[0010] In some possible implementations, the multi-scale pixel-level feature vectors extracted based on the tab mask are extracted pixel-by-pixel or in small blocks. The multi-scale pixel-level feature vectors include multi-color space feature components, texture and local gradient features, higher-order features, spatial information features and / or depth information features.

[0011] In some possible implementations, during the training of the unsupervised probabilistic GMM model, the input is the pixel feature vector in the normal electrode images collected during the training phase, and the output is the probability distribution parameters of the pixel features. When detecting using the unsupervised probabilistic GMM model, the probability distribution parameters of each pixel feature vector to be detected are calculated under the unsupervised probabilistic GMM model. Pixel-level anomalies are identified based on a set anomaly threshold, and temporal filtering and spatial consistency constraints are used to correct the pixel-level anomalies to obtain the corrected pixel-level anomalies.

[0012] In some possible implementations, the pixel-level anomalies are analyzed to obtain candidate anomaly regions, and parameter indices of the candidate anomaly regions are obtained, including: The pixel-level anomalies are clustered using connected components or DBSCAN to obtain several candidate anomaly regions. For each of the candidate anomaly regions, a parameter index is calculated, which includes one or more parameters, specifically: The area of ​​the candidate abnormal region and the proportion of the area of ​​the candidate abnormal region to the electrode region; The perimeter, circularity, and / or aspect ratio of the candidate abnormal region; The minimum distance between the candidate abnormal region and the edge of the electrode tab or the solder joint; The maximum / average height difference in the candidate abnormal region is compared with the surrounding normal region. The persistence of the candidate abnormal region over time.

[0013] In some possible implementations, it is determined whether the parameter index meets the set electrode anomaly judgment rules. If it does, the electrode is determined to be abnormal, and the electrode anomaly result is output, including: If any one or combination of the parameter indicators of the candidate abnormal region satisfies the set electrode abnormality judgment rule, it is determined to be an electrode abnormality, and the electrode abnormality result is output. The electrode abnormality result includes the output candidate box, electrode abnormality type, severity level and coordinate information. The electrode abnormality type includes folding, wrinkling, upturning, misalignment and / or tearing.

[0014] Secondly, the present invention also provides an apparatus for detecting tab abnormalities during the battery cell manufacturing process, comprising: The mask generation unit segments the image of the tab area collected during the battery cell production process to generate a tab mask; The pixel anomaly identification unit extracts multi-scale pixel-level feature vectors based on the tab mask and inputs the multi-scale pixel-level feature vectors into an unsupervised probabilistic GMM model to identify pixel-level anomalies. An anomaly region analysis unit analyzes the pixel-level anomalies to obtain candidate anomaly regions and obtains the parameter indicators of the candidate anomaly regions. The electrode abnormality judgment unit determines whether the parameter index meets the set electrode abnormality judgment rules. If it does, the electrode abnormality is judged and the electrode abnormality result is output.

[0015] Thirdly, the present invention also provides an electronic device, characterized in that it includes: at least one processor; and a memory communicatively connected to the processor; wherein the memory stores instructions executable by the processor, the instructions being executed by the processor to enable the processor to perform the method described thereon.

[0016] Fourthly, the present invention also provides a computer-readable storage medium for storing one or more programs, said one or more programs including computer instructions for causing a computer to perform the method.

[0017] Because the present invention adopts the above technical solution, it has the following characteristics: 1. A multi-level detection strategy that combines the SAM (Segment Anything Model) model and the unsupervised probabilistic GMM (Gaussian Mixture Model).

[0018] This invention is the first to combine the SAM model with the unsupervised probabilistic GMM model. By combining image segmentation with unsupervised learning, it significantly improves the detection accuracy of minute defects (such as foil leakage) in electrode sheets. The SAM model achieves efficient and accurate segmentation of electrode sheet regions, while the unsupervised probabilistic GMM model improves the detection capability of abnormal pixels by modeling the probability distribution of pixel features, thus optimizing the limitations of traditional methods in handling complex anomalies.

[0019] 2. Unsupervised learning reduces manual annotation.

[0020] This invention employs the unsupervised learning method of GMM, which can effectively detect abnormal pixels without requiring a large amount of manually labeled data. This effectively solves the problem that deep learning models usually require a large amount of labeled data, greatly reduces the cost of manual labeling in the production process, and improves the operability and flexibility of the model in practical industrial applications.

[0021] 3. Enhanced defect classification and grading capabilities.

[0022] This invention can not only identify different types of abnormalities such as tab folding, wrinkling, and misalignment, but also classify them according to the severity of the defects, providing more accurate fault location and quality decision support for the production line, and ensuring that corrective measures can be taken quickly when tab abnormalities occur.

[0023] 4. Geometric feature analysis and deep information fusion.

[0024] In the process of abnormal pixel detection and abnormal region clustering, this invention adopts geometric feature analysis technology, which not only identifies two-dimensional color and texture features, but also combines three-dimensional information such as depth images or normal vectors to further improve the ability to judge the phenomena of tab folding, wrinkling and upturning. Especially in the three-dimensional morphological recognition of tab folding, this method has higher accuracy than traditional two-dimensional image analysis methods.

[0025] 5. Adapt to diverse electrode materials and process changes.

[0026] This invention can effectively handle different types of battery cell tabs, adapting to changes in tab materials (copper, aluminum, etc.), color, coating, and surface texture. It requires no manual recalibration and is applicable not only to existing production lines but also to rapid deployment and adaptation under different process and material changes. Furthermore, by employing low-latency back-end processing and threshold strategies, it can achieve real-time alarms and automated linkage at the production line level.

[0027] In summary, this invention combines the adaptability of unsupervised learning with limited samples, the accuracy of segmentation models in region extraction, and the advantages of geometric analysis in industrial interpretability and rule constraints. It can detect tab anomalies in battery cell production with high accuracy and low latency, even with limited abnormal samples, and can identify minute folds, wrinkles, and upturned areas in high-resolution images. It relies minimally on manual annotation, possesses unsupervised or weakly supervised capabilities, can be deployed in real-time on production lines (edge ​​devices or industrial PCs), and provides structured outputs that are easy for automated control systems to use. It also has good generalization ability, adapting to different tab materials, colors, and process variations, and can be widely applied to the detection of tab folds, wrinkles, breaks, misalignments, and other anomalies on battery cell production lines. Attached Figure Description

[0028] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts. In the drawings: Figure 1 This is a flowchart of a method for detecting abnormal tabs during battery cell production, according to an embodiment of the present invention.

[0029] Figure 2 This is a structural diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0030] It should be understood that the terminology used herein is for the purpose of describing particular exemplary embodiments only and is not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms “a,” “an,” and “described” as used herein may also include the plural forms. The terms “comprising,” “including,” “containing,” and “having” are inclusive and therefore indicate the presence of the stated features, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, elements, components, and / or combinations thereof. The method steps, processes, and operations described herein are not construed as requiring them to be performed in a particular order described or illustrated unless the order of performance is explicitly indicated. It should also be understood that additional or alternative steps may be used.

[0031] Although terms such as first, second, third, etc., may be used in this document to describe multiple elements, components, regions, layers, and / or segments, these elements, components, regions, layers, and / or segments should not be limited by these terms. These terms may be used only to distinguish one element, component, region, layer, or segment from another. Unless the context clearly indicates otherwise, terms such as "first," "second," and other numerical terms used herein do not imply order or sequence. Therefore, the first element, component, region, layer, or segment discussed below may be referred to as the second element, component, region, layer, or segment without departing from the teachings of the exemplary embodiments.

[0032] For ease of description, spatial relative terms may be used in the text to describe the relationship of one element or feature relative to another element or feature as shown in the figure. These relative terms include, for example, "inside," "outside," "middle," "outer," "below," "above," etc. Such spatial relative terms are intended to include different orientations of the device in use or operation, other than those depicted in the figure.

[0033] Addressing the numerous problems in existing technologies for tab anomaly detection, this invention provides a method, apparatus, equipment, and medium for tab anomaly detection during battery cell production. The method includes: segmenting an image of the tab region acquired during battery cell production to generate a tab mask; extracting multi-scale pixel-level feature vectors based on the tab mask and inputting these vectors into an unsupervised probabilistic GMM model to identify pixel-level anomalies; analyzing the pixel-level anomalies to obtain candidate anomaly regions and calculating their geometric parameters; determining whether the geometric parameters meet predefined rules, and if so, identifying the anomaly as a tab, and outputting the anomaly result. Therefore, this invention effectively detects abnormal pixels without requiring a large amount of manually labeled data, effectively solving the problem that deep learning models typically require a large amount of labeled data, greatly reducing the manual labeling cost in the production process, and improving the operability and flexibility of the model in practical industrial applications.

[0034] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the invention and to fully convey the scope of the invention to those skilled in the art.

[0035] Example 1: As Figure 1As shown, the method for detecting tab anomalies during battery cell production provided in this embodiment employs a multi-level detection process: image segmentation, feature modeling and abnormal pixel detection, clustering and geometric analysis, and post-processing output. This includes: S1. Image segmentation: The image of the tab area collected during the cell production process is segmented to generate an accurate tab mask, ensuring that subsequent processing is carried out only in the tab area, saving computing resources and avoiding background interference.

[0036] In this embodiment, images of the tab area during the battery cell production process are pre-captured. The image acquisition and triggering are synchronized with the production cycle, ensuring that at least one clear image is captured for the tab area of ​​each battery cell. The specific process is as follows: 1) Use a high-resolution industrial camera to capture images of the tab area, and if necessary, use a ring light or polarized light source to suppress metal reflections. The resolution of the industrial camera can be selected according to the actual situation and is not limited here. It can be selected based on the production line speed and field of view, such as 2048×1536 or 4096×3000.

[0037] 2) Use structured light or binocular stereo cameras to acquire point cloud or depth images of the tab region to determine the tab uplift height and wrinkle depth.

[0038] In this embodiment, the image of the electrode area collected during the battery cell production process is segmented to generate an accurate electrode mask. Specifically, a lightweight segmentation network and a pre-trained SAM model are used for segmentation. Segmentation only needs to locate the approximate outline of the electrode and generate a mask, avoiding pixel-level binarization at this stage to preserve the flexibility of subsequent discrimination. To improve speed, the image segmentation in this embodiment can adopt a two-stage strategy, the specific process of which is as follows: First, a lightweight segmentation network is used to quickly locate the large bounding box of the electrode (the large bounding box of the electrode refers to the outer contour of the electrode) to obtain a sub-image. The lightweight segmentation network can be a small YOLOv11 model, etc., which is used as an example, but is not limited to this.

[0039] Then, the pre-trained SAM model is used to segment the above subgraph to obtain the pole ear mask, thereby reducing the amount of computation while ensuring accuracy.

[0040] S2. Feature Modeling and Abnormal Pixel Detection: Extract multi-scale pixel-level feature vectors within the tab mask and input them into an unsupervised probabilistic GMM model to identify pixel-level abnormal points.

[0041] In this embodiment, multi-scale pixel-level feature vectors are extracted within the tab mask. The specific process is as follows: S21. Within the segmented tab mask, extract the following pixel-level feature vectors pixel-by-pixel or in small blocks, including: color space, texture and local gradient, higher-order features, spatial information and / or depth information, wherein: Color spaces: RGB, HSV, CIE-Lab and other color space components.

[0042] Textures and local gradients: Sobel operator, Laplacian operator, Local Binary Pattern (LBP), Laplacian of Gaussian operator, etc.

[0043] Advanced features such as local contrast, local variance, and Laplace response are useful for capturing high-frequency signals that contribute to wrinkle formation. Spatial Information: Pixel coordinates are normalized to (x, y) to obtain the relative position of the pixel to the tab main axis / solder joint. Specifically: Pixel coordinate normalization (x, y): The original coordinates of the pixel (e.g., xx in the image width direction and yy in the height direction) are divided by the image width WW and height HH respectively to obtain normalized coordinates (x / W, y / H), ensuring that the coordinate range falls within the interval and eliminating the influence of image size differences; Relative position of the pixel to the tab main axis / solder joint: First, the tab main axis and the solder joint are obtained through image processing methods, such as using Hough transform to detect straight lines to obtain the tab main axis, and using template matching or feature point detection to locate the solder joint; Then, the spatial relative relationship between the pixel coordinates and the main axis (e.g., the distance from the pixel to the main axis line, the offset along the main axis direction) or the solder joint (e.g., the Euclidean distance between the pixel and the solder joint, the difference in direction angle).

[0044] Depth information includes depth values ​​and local normal vectors, used to identify upturn height or depth wrinkles. Specifically: Depth information is defined as the distance from the object's surface corresponding to a pixel to the sensor (or reference plane) in the case of a 3D image (such as a depth map or point cloud projection); local normal vectors are defined as the direction vector perpendicular to the local surface of the object. The normal vector is calculated by fitting a plane of local neighboring points (such as the k-nearest neighbors of a 3D point cloud) to describe the orientation of the local surface and identify upturn height or depth wrinkles.

[0045] Furthermore, pixel-by-pixel extraction directly extracts the feature vectors at the pixel level, centered on a single pixel. Extraction in small blocks uses windows such as 7×7 or 11×11 to extract the pixel-level feature vectors within the window.

[0046] Furthermore, for different regions within the tab (such as near the solder joint, the distal end, and the edge), a partitioning strategy can be used to extract pixel-level feature vectors to capture local differences, where: The area near the solder joint typically exhibits high brightness variations and complex texture structures, making it suitable for extracting high-frequency texture features and edge response features. The distant region has relatively uniform features and is more suitable for extracting color distribution and low-frequency texture features; Edge regions: These regions exhibit clear boundary transitions and are suitable for extracting gradient direction and edge intensity features.

[0047] S22. In order to reduce the modeling complexity of the unsupervised probabilistic GMM model, the above pixel-level feature vectors can be normalized and PCA reduced in dimensionality as needed.

[0048] S23. Unsupervised probabilistic GMM model modeling, the process is as follows: S231. During the training or initialization phase, the unsupervised probabilistic GMM model is trained using only normal electrode data (a large number of normal samples are relatively easy to collect). The training process of the unsupervised probabilistic GMM model is as follows: the input is the pixel feature vector in the collected normal electrode image, and the output is the probability distribution parameters of the pixel feature vector.

[0049] S232. In the detection phase, the probability distribution parameters of each pixel feature vector to be detected under the unsupervised probabilistic GMM model are calculated, and pixel-level outliers are obtained based on the set outlier threshold. The outlier threshold can be automatically set according to the robust statistics of the training set, for example, the negative log-likelihood of the 95th percentile of the training set can be used as the outlier threshold, but it is not limited to this.

[0050] Furthermore, to avoid misjudgments caused by isolated noise points, temporal filtering and spatial consistency constraints are used to correct the obtained pixel-level outliers, resulting in corrected pixel-level outliers, where: The process of steady-state detection of abnormal pixels in consecutive frames using temporal filtering is as follows: the real defects of battery tabs on the production line usually appear stably in multiple consecutive frames, while random noise does not. When the input is a pixel-level abnormal point (0 for normal, 1 for abnormal) detected by an unsupervised probabilistic GMM model in N consecutive frames, the first step is to use logical judgment, such as counting: for a certain pixel position (x, y) in the current frame, check how many frames it has been marked as abnormal in the past N frames; then a threshold judgment is performed: a temporal threshold Tt is set, and only when the number of times the pixel is judged as abnormal in N frames is greater than or equal to Tt is the final output judged as abnormal; otherwise, it is corrected to normal.

[0051] Spatial consistency constraints were not directly used in the modeling process, but a consistency assessment of neighboring pixels was introduced during the detection phase to assist decision-making. For example, if a pixel is judged as abnormal, but most of its surrounding pixels are normal, it may be corrected to a normal pixel, and vice versa. This constraint helps reduce misjudgments caused by local noise and enhances the spatial continuity of the detection results.

[0052] S3. Clustering and Geometric Analysis: Density clustering (such as DBSCAN) or connected component analysis is used to merge the scattered points into candidate anomaly regions and calculate the parameter indices of the candidate anomaly regions.

[0053] In this embodiment, density clustering (such as DBSCAN) or connected component analysis is used to merge the identified pixel anomalies into candidate anomaly regions, and the parameter indicators of the candidate anomaly regions are calculated. The specific process is as follows: S31. Cluster the pixel anomaly points according to connected components or DBSCAN to obtain several candidate anomaly regions.

[0054] S32. Analyze each candidate anomaly region to obtain parameter indicators, wherein the parameter indicators may include: The area (number of pixels) of the candidate anomaly region and the proportion of the candidate anomaly region to the tab region; Shape description parameters of candidate anomaly regions: perimeter, roundness, stretch ratio, etc. of candidate anomaly regions; Minimum distance between candidate anomaly region and tab edge or solder joint: For example, if the candidate anomaly region is pixel set A and the tab edge / solder joint is contour or region set E, the minimum distance between the candidate anomaly region and tab edge or solder joint is the shortest distance from each pixel in set A to each pixel in set E. If a depth map is available, calculate the maximum / average height difference in the candidate abnormal region and compare it with the surrounding normal region. Persistence in the temporal dimension (whether it persists across multiple frames).

[0055] S4. Post-processing output: Based on the set electrode abnormality judgment rules, any one or combination of parameter indicators is judged. If the conditions are met, the electrode abnormality is judged and the electrode abnormality result is output.

[0056] In this embodiment, the rule for judging abnormalities of the electrode tab is as follows: Suppose we want to judge whether a battery electrode tab has "edge damage" or "foreign object" abnormalities: If (the area of ​​the candidate region accounts for more than 1% of the total area of ​​the electrode tab) and (the minimum distance between the candidate region and the edge of the electrode tab is 0 pixels, i.e., close to the edge) and (the abnormality exists in 3 consecutive frames), then it is judged as "edge damage of the electrode tab". This is an example, but it is not limited to this. The rules can be set as needed.

[0057] In this embodiment, the tab anomaly results include candidate boxes, anomaly types (folding, wrinkling, warping, misalignment, tearing), severity levels, and coordinate information, which are used by the production line control system. Specifically: Mark abnormal areas on the original image with a rectangle or precise mask, and output a detection report in JSON format: {type, top left coordinate, bottom right coordinate, area, severity, suggested action (manual re-inspection / removal / shutdown)}.

[0058] Linking anomalies to production time-series data (barcodes / serial numbers) facilitates early warning, statistics, and traceability.

[0059] It supports online threshold adjustment and offline playback analysis, facilitating engineering optimization.

[0060] Furthermore, severity grading can be based on multi-dimensional thresholds such as area, shape, or positional deviation, and the risk level can be quantified through methods such as weighted scoring or decision trees. It can also be adjusted according to the actual requirements of the production line to output the final severity grading (mild / moderate / severe), which is determined according to actual production needs and will not be elaborated here.

[0061] The effectiveness of the method for detecting tab abnormalities during battery cell production provided by the present invention will be verified in detail through specific embodiments below.

[0062] This embodiment compares the results with YOLOv8n (RGB-supervised training only, 500 fine-tuning iterations) on a production line dataset, running on machines with the same configuration. Table 1 shows the comparison results of tab fold detection. Normal samples account for 99% of the dataset, while abnormal samples are scarce and mostly minor folds.

[0063] Table 1 Comparison results of electrode folding test

[0064] In summary, compared with traditional methods and purely supervised deep models, this invention has the following advantages: 1) Advantage of fewer samples: This invention significantly reduces the need for anomaly labeled data by using unsupervised modeling with normal samples as the main component, making it easier to promote on real production lines.

[0065] 2) Strong microstructure recognition capability: The pixel-level modeling and multimodal features (color + gradient + depth) of this invention are combined to effectively identify folds / creases with extremely small width and height.

[0066] 3) Industrial-grade real-time performance: This invention adopts ROI pre-screening, hierarchical discrimination and vectorization to meet the production line cycle time requirements and has online alarm capabilities.

[0067] 4) Explainable and traceable: This invention provides an easily understandable basis for anomaly determination through geometric features and rule constraints, which facilitates engineers in locating the cause and improving the process.

[0068] 5) Easy to integrate and expand: The present invention adopts a modular design of segmentation, modeling, clustering, and output, which facilitates the interface with existing production line systems (MES / PLC) and supports the subsequent introduction of self-supervised and few-shot learning modules.

[0069] Example 2: Following the method for detecting tab anomalies during battery cell production provided in Example 1, this example provides an apparatus for detecting tab folding anomalies during battery cell production. The apparatus provided in this example can implement the method for detecting tab anomalies during battery cell production as described in Example 1. This apparatus can be implemented through software, hardware, or a combination of both. For ease of description, this example is described by dividing the functions into various units. Of course, in implementation, the functions of each unit can be implemented in one or more software and / or hardware components. For example, the apparatus may include integrated or separate functional modules or units to execute the corresponding steps in the methods of Example 1. Since the apparatus in this example is basically similar to the method example, the description process of this example is relatively simple. Relevant details can be found in the description of Example 1. The embodiment of the apparatus for detecting tab folding anomalies during battery cell production provided by this invention is merely illustrative.

[0070] Specifically, the device for detecting tab abnormalities during battery cell production provided in this embodiment includes: The mask generation unit segments the image of the tab area collected during the battery cell production process to generate a tab mask; The pixel anomaly identification unit extracts multi-scale pixel-level feature vectors based on the tab mask and inputs the multi-scale pixel-level feature vectors into an unsupervised probabilistic GMM model to identify pixel-level anomalies. An anomaly region analysis unit analyzes the pixel-level anomalies to obtain candidate anomaly regions and obtains the parameter indicators of the candidate anomaly regions. The electrode abnormality judgment unit determines whether the parameter index meets the set electrode abnormality judgment rules. If it does, the electrode abnormality is judged and the electrode abnormality result is output.

[0071] Example 3: This example provides an electronic device corresponding to the method for detecting abnormal electrode tabs during the cell production process provided in Example 1. The electronic device can be an electronic device for the client, such as a mobile phone, laptop, tablet computer, desktop computer, etc., to execute the method of Example 1.

[0072] like Figure 2As shown, the electronic device includes a processor, a memory, a communication interface, and a bus. The processor, memory, and communication interface are connected via the bus to enable communication between them. The memory stores a computer program that can run on the processor. When the processor runs the computer program, it executes the method of Embodiment 1. The implementation principle and technical effects are similar to those of Embodiment 1, and will not be repeated here. Those skilled in the art will understand that... Figure 2 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computing device on which the present application is applied. The specific computing device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0073] In a preferred embodiment, the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, and can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), and optical discs.

[0074] In a preferred embodiment, the processor can be any type of general-purpose processor such as a central processing unit (CPU) or a digital signal processor (DSP), and is not limited thereto.

[0075] Example 4: This example provides a computer-readable storage medium for storing one or more programs, the one or more programs including computer instructions, which, when executed by a computer, cause the computer to perform the method provided in Example 1 above.

[0076] In a preferred embodiment, the computer-readable storage medium may be a tangible device for holding and storing instructions executable, such as, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any combination thereof. The computer-readable storage medium stores computer program instructions that cause a computer to perform the method provided in Embodiment 1 above.

[0077] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (devices), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0078] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0079] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0080] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In the description of this specification, the terms "a preferred embodiment," "furthermore," "specifically," "in this embodiment," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the embodiments in this specification. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described can be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting abnormal tabs during battery cell production, characterized in that, include: The image of the tab area collected during the battery cell production process is segmented to generate a tab mask; Based on the aforementioned tab mask, multi-scale pixel-level feature vectors are extracted, and the multi-scale pixel-level feature vectors are input into an unsupervised probabilistic GMM model to identify pixel-level outliers. The pixel-level anomalies are analyzed to obtain candidate anomaly regions, and the parameter indices of the candidate anomaly regions are obtained. Determine whether the parameter index meets the set electrode abnormality judgment rules. If it does, the electrode is judged to be abnormal and the electrode abnormality result is output.

2. The method for detecting abnormal tabs during battery cell production according to claim 1, characterized in that, It also includes the step of acquiring images of the tab area during the cell manufacturing process, the process of which is as follows: Use an industrial camera to collect data on the electrode area; Collect point cloud or depth images of the electrodes.

3. The method for detecting abnormal tabs during battery cell production according to claim 2, characterized in that, The process of segmenting images of the tab area acquired during battery cell production to generate a tab mask includes: A lightweight segmentation network is used to locate the large bounding box of the pole ears in the image of the pole ear region to obtain a sub-image; The sub-image is segmented using a pre-trained SAM model to obtain the pole ear mask.

4. The method for detecting abnormal tabs during battery cell production according to claim 1, characterized in that, The multi-scale pixel-level feature vectors extracted based on the aforementioned electrode mask are extracted pixel-by-pixel or in small blocks. The multi-scale pixel-level feature vectors include multi-color space feature components, texture and local gradient features, higher-order features, spatial information features, and / or depth information features.

5. The method for detecting abnormal tabs during battery cell production according to claim 1, characterized in that, When training an unsupervised probabilistic GMM model, the input is the pixel feature vector in the normal electrode images collected during the training phase, and the output is the probability distribution parameters of the pixel features. When detecting using an unsupervised probabilistic GMM model, the probability distribution parameters of each pixel feature vector to be detected are calculated under the unsupervised probabilistic GMM model. Pixel-level anomalies are identified based on a set anomaly threshold, and temporal filtering and spatial consistency constraints are used to correct the pixel-level anomalies to obtain the corrected pixel-level anomalies.

6. The method for detecting abnormal tabs during battery cell production according to claim 5, characterized in that, Analyzing the pixel-level anomalies yields candidate anomaly regions, and the parameter indices of these candidate anomaly regions are obtained, including: The pixel-level anomalies are clustered using connected components or DBSCAN to obtain several candidate anomaly regions. For each of the candidate anomaly regions, a parameter index is calculated, which includes one or more parameters, specifically: The area of ​​the candidate abnormal region and the proportion of the area of ​​the candidate abnormal region to the electrode region; The perimeter, circularity, and / or aspect ratio of the candidate abnormal region; The minimum distance between the candidate abnormal region and the edge of the electrode tab or the solder joint; The maximum / average height difference in the candidate abnormal region is compared with the surrounding normal region. The persistence of the candidate abnormal region over time.

7. The method for detecting abnormal tabs during battery cell production according to claim 6, characterized in that, Determine whether the parameter indicators meet the set electrode abnormality judgment rules. If they do, the electrode is determined to be abnormal, and the electrode abnormality result is output, including: If any one or combination of the parameter indicators of the candidate abnormal region satisfies the set electrode abnormality judgment rule, it is determined to be an electrode abnormality, and the electrode abnormality result is output. The electrode abnormality result includes the output candidate box, electrode abnormality type, severity level and coordinate information. The electrode abnormality type includes folding, wrinkling, upturning, misalignment and / or tearing.

8. A device for detecting abnormal tabs during battery cell production, characterized in that, include: The mask generation unit segments the image of the tab area collected during the battery cell production process to generate a tab mask; The pixel anomaly identification unit extracts multi-scale pixel-level feature vectors based on the tab mask and inputs the multi-scale pixel-level feature vectors into an unsupervised probabilistic GMM model to identify pixel-level anomalies. An anomaly region analysis unit analyzes the pixel-level anomalies to obtain candidate anomaly regions and obtains the parameter indicators of the candidate anomaly regions. The electrode abnormality judgment unit determines whether the parameter index meets the set electrode abnormality judgment rules. If it does, the electrode abnormality is judged and the electrode abnormality result is output.

9. An electronic device, characterized in that, include: At least one processor; And a memory communicatively connected to the processor; wherein the memory stores instructions executable by the processor, the instructions being executed by the processor to enable the processor to perform the method according to any one of claims 1-7.

10. A computer-readable storage medium for storing one or more programs, characterized in that, The one or more programs include computer instructions for causing a computer to perform the method according to any one of claims 1-7.

Images