Coating foil leakage detection method and device in battery cell production process

Abstract

The invention discloses a coating foil leakage detection method and device in a battery cell production process, and belongs to the technical field of battery production, and the detection method comprises the steps: obtaining a pole piece image in the battery cell production process; segmenting the pole piece image by using a pre-trained SAM model, and extracting a pole piece region to obtain a mask image; pixel feature extraction is carried out on the mask image; identifying the extracted pixel features by adopting a pre-trained GMM (Gaussian Mixture Model), and distinguishing normal pixel points from abnormal pixel points; and carrying out clustering analysis on the abnormal pixel points to determine a foil leakage area. According to the method, an advanced image segmentation model (SAM) and a Gaussian mixture model (GMM) are combined to carry out unsupervised learning detection, clustering analysis of abnormal pixel points is matched, and industrial-grade real-time detection with excellent performance is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of battery production, and particularly to a method and device for detecting coating foil leakage during the production process of battery cells. Background Art

[0002] The electrode sheet is an important component of the battery, and its performance directly affects the overall performance and lifespan of the battery. Therefore, quality control of the electrode sheet is crucial, especially for the detection of abnormalities in the electrode sheet. During the manufacturing process of lithium batteries, improper production processes may lead to abnormal bonding of the electrode sheet, and such batteries often fail to meet production standards. In recent years, with the development of deep learning technology, some studies have begun to attempt to apply deep learning to the detection of electrode sheet abnormalities. For example, by using a convolutional neural network (CNN) to extract features from the electrode sheet image, and then using traditional machine learning models such as a support vector machine (SVM) or a random forest (Random Forest) for classification to determine whether the electrode sheet is normal. During the production process of the electrode sheet, the integrity of the electrode sheet is directly related to battery performance and safety. However, the phenomenon of foil leakage (i.e., there are small defect areas or uncovered parts on the surface of the electrode sheet) is difficult to accurately identify through traditional visual detection methods. The problems existing in traditional visual detection methods mainly include: Image segmentation technology: It can segment the electrode sheet area, but its ability to detect abnormal pixel points is insufficient.

[0003] Deep learning methods: Although they can be used for detection, their ability to recognize subtle features of the foil leakage phenomenon is weak. The features of the foil leakage phenomenon are subtle, usually small in size, and may not reach the minimum recognizable unit of the target detection algorithm. This results in poor detection effects of traditional convolutional neural networks on the foil leakage phenomenon, especially in the recognition of fine features of high-resolution images.

[0004] Difficulty in data marking: Existing deep learning-based abnormality detection methods usually require a large amount of marked data to train the model. However, in the actual application of electrode sheet abnormality detection, the number of abnormal samples is often very small, making this method difficult to implement. The abnormal situations of the electrode sheet are mostly difficult to detect due to similar colors, which brings certain difficulties to model recognition.

[0005] Limited ability to recognize complex and non-linear abnormal patterns: Traditional machine vision technologies or shallow learning models can only process simple and regular image features, and have limited ability to recognize complex and non-linear abnormal patterns. Deep learning models designed for specific types and characteristics of electrode sheets may not be able to generalize well to other types of electrode sheets. Summary of the Invention

[0006] The object of the present invention is to overcome the deficiencies in the prior art and provide a method and device for detecting coating foil leakage during the production of electric cores, so as to solve the technical problems that the traditional visual detection method cannot accurately identify coating foil leakage and has poor generalization ability and adaptability.

[0007] To achieve the above object, the present invention is implemented by the following technical solutions: In the first aspect, the present invention provides a method for detecting coating foil leakage during the production of electric cores, including: Obtaining a pole piece image during the production of the electric core; Segmenting the pole piece image and extracting the pole piece region to obtain a mask image; Extracting pixel features from the mask image; Identifying the extracted pixel features to distinguish normal pixel points and abnormal pixel points; Performing clustering analysis on the abnormal pixel points to determine the foil leakage region.

[0008] Optionally, the segmenting the pole piece image includes: using a pre-trained SAM model for image segmentation.

[0009] The SAM (Segment Anything Model) model is a general image segmentation model developed by Meta AI, with zero-shot generalization ability, and can achieve the segmentation of any object through various forms of prompts (such as points, boxes, texts). The SAM model consists of three core components: Image Encoder: Using Vision Transformer (ViT) or a convolutional network, encoding the input image into high-dimensional features. For example, for an image with a resolution of 1024×1024, ViT divides it into 16×16 Patches and generates a feature map of 64×64×768.

[0010] Prompt Encoder: Processing the prompt information (such as points, boxes, texts) input by the user and converting it into an embedding vector compatible with the image features. The text prompt is encoded by the CLIP model, and the spatial prompts (points, boxes) use positional encoding.

[0011] Mask Decoder: Fusing the image features and the prompt embedding, generating the final segmentation mask through a Transformer structure, and supporting multi-mask output to handle ambiguity (such as partial and overall segmentation).

[0012] Optionally, the identifying the extracted pixel features includes: using a pre-trained GMM model for image recognition.

[0013] The GMM (Gaussian Mixture Model) is a probabilistic clustering model that assumes data is generated by a linear combination of multiple Gaussian distributions and can fit data with arbitrarily complex distributions.

[0014] The estimation of the GMM model relies on the Expectation-Maximization (EM) algorithm, which iterates in two stages: E-step (expectation): calculates the "responsibility value" (posterior probability) of each data point belonging to each Gaussian distribution; M-step (maximization): updates the model parameters.

[0015] In summary, the present invention precisely segments the pole piece image through the SAM model, effectively extracts the pole piece area, and combines the GMM model to achieve precise detection of abnormal pixel points. This combined method significantly improves the accuracy of real-time detection and can quickly and accurately identify pole piece abnormalities in the production line. It can effectively handle abnormal scenarios of different types of battery cell pole pieces, including but not limited to uneven coating, surface foil leakage, and other minor defects, meeting the diverse industrial production requirements.

[0016] Optionally, the clustering analysis of the abnormal pixel points includes: Using a density clustering algorithm to group adjacent abnormal pixel points; Taking the abnormal pixel point area generated by each group as the clustering cluster of the candidate area; Obtaining the number of pixel points in each clustering cluster; If the number of pixel points exceeds a preset number threshold, the corresponding candidate area is a foil leakage area.

[0017] In summary, through the unsupervised learning ability of the GMM model and combined with the clustering analysis algorithm of abnormal pixel points (such as DBSCAN), the present invention can efficiently identify minor defect areas such as foil leakage. This method significantly reduces the missed detection rate of minor defects on the production line, ensuring product quality and production efficiency. Traditional deep learning detection methods rely on a large amount of manually labeled data, while the present invention uses unsupervised learning technology, which can complete the detection task through automated modeling and analysis under the condition of limited abnormal samples, significantly reducing the workload and cost of manual labeling.

[0018] Optionally, after determining the foil leakage area, it further includes: Marking the foil leakage area on the pole piece image and generating a detection report; the detection report includes the marked pole piece image and the information parameters of the foil leakage area.

[0019] By generating a detection report, the present invention can achieve a real-time feedback function, can be integrated with the existing production line system, and realize instant monitoring of each production cycle and output of quality reports. The detection results can be directly transmitted to the production control system to achieve automated and intelligent quality control.

[0020] In a second aspect, the present invention provides a coating leakage detection device during the production of battery cells, comprising: An image acquisition module configured to acquire the image of the electrode sheet during the production of battery cells; An image segmentation module configured to segment the electrode sheet image and extract the electrode sheet area to obtain a mask image; A feature extraction module configured to extract pixel features from the mask image; A pixel recognition module configured to recognize the extracted pixel features and distinguish normal pixel points from abnormal pixel points; A leakage determination module configured to perform cluster analysis on the abnormal pixel points to determine the leakage area.

[0021] Optionally, the segmentation of the electrode sheet image includes: performing image segmentation using a pre-trained SAM model; the recognition of the extracted pixel features includes: performing image recognition using a pre-trained GMM model.

[0022] In summary, the present invention combines an advanced image segmentation model (SAM) and a Gaussian mixture model (GMM) for unsupervised learning detection, and cooperates with the cluster analysis of abnormal pixel points to achieve superior performance in real-time industrial coating leakage detection.

[0023] In a third aspect, the present invention provides an electronic device, comprising a processor and a storage medium; The storage medium is used to store instructions; The processor is used to operate according to the instructions to execute the steps of the above method.

[0024] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the steps of the above method are implemented.

[0025] In a fifth aspect, the present invention provides a computer program product, comprising a computer program / instructions, and when the computer program / instructions are executed by a processor, the steps of the above method are implemented.

[0026] Compared with the prior art, the beneficial effects achieved by the present invention: A method and device for detecting coating leakage of foils during the production process of battery cells. The present invention combines the SAM model with the GMM model, and through the combination of image segmentation and unsupervised learning, significantly improves the detection accuracy of tiny defects (such as foil leakage) in the electrode sheet. The SAM model realizes the efficient and accurate segmentation of the electrode sheet area, while the Gaussian mixture model improves the detection ability of abnormal pixels by modeling the probability distribution of pixel features, and optimizes the limitations of traditional methods in dealing with complex anomalies. The GMM model uses an unsupervised learning method and can effectively detect abnormal pixels without a large amount of manually labeled data. It effectively solves the problem that deep learning models usually require a large amount of labeled data, greatly reduces the manual labeling cost in the production process, and improves the operability and flexibility of the model in actual industrial applications.

[0027] The detection method of the present invention can perform anomaly detection on the electrode sheets on the production line in real time, and has a high processing speed and real-time feedback ability. This solution can be seamlessly integrated with existing production equipment, support large-scale industrial applications, and at the same time ensure the accuracy of the detection results, meeting the high requirements for detection efficiency and accuracy in the industrial production process. At the same time, the present invention has good generalization ability and can be extended to the quality detection of different production lines. Description of the Drawings

[0028] Figure 1 It is a schematic flowchart of the method for detecting coating leakage of foils during the production process of battery cells provided by an embodiment of the present invention. Detailed Embodiments

[0029] The present invention will be further described below with reference to the drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and cannot be used to limit the protection scope of the present invention.

[0030] Embodiment 1:

[0031] As Figure 1 shown, an embodiment of the present invention provides a method for detecting coating leakage of foils during the production process of battery cells, including the following steps: Step S1, obtain the electrode sheet image during the production process of the battery cell.

[0032] Specifically, in this embodiment, the electrode sheet image during the production process of the battery cell is obtained through a high-resolution camera or sensor. Ensure that the image quality meets the requirements of subsequent processing, including the surface characteristics of the electrode sheet and possible defect areas.

[0033] Step S2, segment the electrode sheet image and extract the electrode sheet area to obtain a mask image.

[0034] Specifically, in this embodiment, segmenting the electrode sheet image includes: using a pre-trained SAM model for image segmentation.

[0035] The SAM (Segment Anything Model) is a general-purpose image segmentation model developed by Meta AI. It has zero-shot generalization ability and can segment any object through various forms of prompts (such as points, boxes, text). The SAM model consists of three core components: Image Encoder: It uses Vision Transformer (ViT) or a convolutional network to encode the input image into high-dimensional features. For example, for an image with a resolution of 1024×1024, ViT divides it into 16×16 patches and generates a feature map of 64×64×768.

[0036] Prompt Encoder: It processes the prompt information (such as points, boxes, text) input by the user and converts it into an embedding vector compatible with the image features. Text prompts are encoded through the CLIP model, and spatial prompts (points, boxes) use positional encoding.

[0037] Mask Decoder: It fuses the image features and the prompt embeddings, and generates the final segmentation mask through a Transformer structure, supporting multi-mask output to handle ambiguity (such as partial and whole segmentation).

[0038] The SAM model is trained based on a large-scale dataset and can efficiently and accurately identify the target regions in the image. The segmentation result generates a pole piece mask for subsequent anomaly detection and analysis. Through the segmentation result of the SAM model, the pole piece region is separated from the background and other interference factors, and a mask image of the pole piece region is obtained. This mask image provides an accurate region boundary for the subsequent steps and avoids misjudgment in traditional visual detection methods when the edges are not clear.

[0039] Step S3: Extract pixel features from the mask image.

[0040] Specifically, in this embodiment, the pixel features include color features (such as RGB values) and spatial position information (such as pixel coordinates). The color features reflect the visual features of the pixels, and the spatial position information helps to analyze the shape and distribution of the abnormal regions.

[0041] Step S4: Identify the extracted pixel features to distinguish normal pixel points and abnormal pixel points.

[0042] Specifically, in this embodiment, identifying the extracted pixel features includes: using a pre-trained GMM model for image recognition.

[0043] The GMM (Gaussian Mixture Model) is a probabilistic clustering model. It assumes that the data is generated by a linear combination of multiple Gaussian distributions and can fit data with any complex distribution.

[0044] The estimation of the GMM model relies on the Expectation-Maximization (EM) algorithm, which iterates in two stages: E-step (expectation): Calculate the "responsibility value" (posterior probability) of each data point belonging to each Gaussian distribution; M-step (maximization): Update the model parameters.

[0045] The GMM differentiates between normal and abnormal pixels by fitting the color features and spatial location information of pixels to multiple Gaussian distributions.

[0046] For each pixel point, calculate the probability that it belongs to the normal component. If the probability value of a certain pixel point is lower than the set threshold, then this pixel point is determined to be an abnormal pixel. Through the GMM model, it is possible to accurately identify minute defects or foil leakage phenomena, even if these defects are difficult to identify in traditional visual inspections.

[0047] In summary, the present invention accurately segments the pole piece image through the SAM model, effectively extracts the pole piece area, and combines the GMM model to achieve precise detection of abnormal pixel points. This combined method significantly improves the accuracy of real-time detection and can quickly and accurately identify pole piece abnormalities in the production line. It can effectively handle abnormal scenarios of different types of battery cell pole pieces, including but not limited to uneven coating, surface foil leakage, and other minute defects, meeting the diverse industrial production requirements.

[0048] Step S5: Perform clustering analysis on the abnormal pixel points to determine the foil leakage area.

[0049] Specifically, in this embodiment, the clustering analysis of the abnormal pixel points includes: Step S5.1: Use a density clustering algorithm to group adjacent abnormal pixel points.

[0050] Density clustering algorithms such as DBSCAN can effectively merge the abnormal pixel points scattered in the image into one or more regions.

[0051] Step S5.2: Take the abnormal pixel point regions generated by each group as the clustering clusters of the candidate regions.

[0052] Step S5.3: Obtain the number of pixel points in each clustering cluster.

[0053] Step S5.4: If the number of pixel points exceeds the preset quantity threshold, then the corresponding candidate region is the foil leakage area.

[0054] The number of pixels calculates the "area" of this region. If the area of a certain clustering cluster exceeds the preset threshold, then this region is considered a foil leakage region. The area threshold can be adjusted according to the requirements of different production lines to adapt to different detection accuracy requirements. For abnormally small regions, they can be filtered out according to specific rules to reduce unnecessary false detections.

[0055] In summary, through the unsupervised learning ability of the GMM model and combined with the clustering analysis algorithm for abnormal pixels (such as DBSCAN), the present invention can efficiently identify tiny defect regions such as foil leakage. This method significantly reduces the missed detection rate of tiny defects on the production line, ensuring product quality and production efficiency. Traditional deep learning detection methods rely on a large amount of manually labeled data, while the present invention uses unsupervised learning technology, which can complete the detection task through automated modeling and analysis under the condition of limited abnormal samples, significantly reducing the workload and cost of manual labeling.

[0056] Optionally, after determining the foil leakage region, it further includes: Mark the foil leakage region on the electrode sheet image and generate a detection report; the detection report includes the marked electrode sheet image and the information parameters of the foil leakage region.

[0057] Specifically, in this embodiment, the marking uses a highlighted rectangular box to prominently display the abnormal region, facilitating the operator to quickly identify and subsequent processing. The information parameters of the foil leakage region include the area, position, detection time, etc. of the foil leakage region. These information contribute to the real-time quality control and data analysis of the production line.

[0058] By generating a detection report, the present invention can achieve a real-time feedback function, which can be integrated with the existing production line system to achieve instant monitoring of each production cycle and output of quality reports. The detection results can be directly transmitted to the production control system to achieve automated and intelligent quality control.

[0059] To verify the detection method provided in this embodiment, in the electrode sheet defect detection, using the official model of the commonly used object detection algorithm YOLOv8 in the industry, under the same dataset and training set, and running the same training times on the same configured machine, Table 1 shows the results. Among them, the YOLOv8n model is provided by the official, and on this basis, after training 500 times as a pre-trained model and comparing with the new method of the team, it is found that our effect is comprehensively higher than the method provided by the official.

[0060] Table 1: Comparison of electrode sheet tiny defect (foil leakage) detection models

[0061] Example Two: The embodiment of the present invention provides a coating foil leakage detection device during the production process of an electric core, including: An image acquisition module, configured to acquire the pole piece images during the production process of the battery cell; An image segmentation module, configured to segment the pole piece images using a pre-trained SAM model, and extract the pole piece region to obtain a mask image; A feature extraction module, configured to extract pixel features from the mask image; A pixel recognition module, configured to recognize the extracted pixel features using a pre-trained GMM model to distinguish normal pixel points and abnormal pixel points; A foil leakage determination module, configured to perform clustering analysis on the abnormal pixel points to determine the foil leakage region.

[0062] In summary, the present invention combines an advanced image segmentation model (SAM) and a Gaussian mixture model (GMM) for unsupervised learning detection, and cooperates with the clustering analysis of abnormal pixel points to achieve superior performance in real-time detection of industrial coating foil leakage.

[0063] Embodiment Three:

[0064] Based on the detection method provided in Embodiment One, an embodiment of the present invention provides an electronic device, including a processor and a storage medium; The storage medium is used to store instructions; The processor is used to operate according to the instructions to execute the steps of the above method.

[0065] Embodiment Four:

[0066] Based on the detection method provided in Embodiment One, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the steps of the above method are implemented.

[0067] Embodiment Five:

[0068] Based on the detection method provided in Embodiment One, an embodiment of the present invention provides a computer program product, including a computer program / instructions, and when the computer program / instructions are executed by a processor, the steps of the above method are implemented.

[0069] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0070] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to produce a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices produce means for implementing the functions specified in one flow Figure 1 one flow or more flows and / or blocks Figure 1 or means for implementing the functions specified in a block or more blocks.

[0071] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory produce a manufactured article including instruction means for implementing the functions specified in one flow Figure 1 one flow or more flows and / or blocks Figure 1 or means for implementing the functions specified in a block or more blocks.

[0072] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one flow Figure 1 one flow or more flows and / or blocks Figure 1 or means for implementing the functions specified in a block or more blocks.

[0073] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the technical principles of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. A method for detecting coating leakage of foil during the production process of an electric core, characterized in that, Including: Obtain the pole piece image during the production process of the battery cell; Segment the pole piece image, and extract the pole piece area to obtain a mask image; Extract pixel features from the mask image; Identify the extracted pixel features to distinguish normal pixel points and abnormal pixel points; Perform clustering analysis on the abnormal pixel points to determine the foil leakage area.

2. The method for detecting coating leakage of the electrode core during production according to claim 1, characterized in that The segmenting the pole piece image includes: performing image segmentation using a pre-trained SAM model.

3. The coating leakage foil detection method during the production process of the battery cell according to claim 1, wherein, The identifying the extracted pixel features includes: performing image recognition using a pre-trained GMM model.

4. The coating leakage foil detection method during the production process of the battery cell according to claim 1, wherein The performing clustering analysis on the abnormal pixel points includes: Using a density clustering algorithm to group adjacent abnormal pixel points; Taking the abnormal pixel point area generated by each group as a clustering cluster of the candidate area; Obtain the number of pixel points in each clustering cluster; If the number of pixel points exceeds a preset number threshold, the corresponding candidate area is the foil leakage area.

5. The method for detecting coating leakage of the electrode core during production according to claim 1, wherein After determining the foil leakage area, it further includes: Mark the foil leakage area on the pole piece image and generate a detection report; the detection report includes the marked pole piece image and the information parameters of the foil leakage area.

6. A coating leakage detection device during the production process of an electric core, characterized in that, Including: An image acquisition module configured to obtain the pole piece image during the production process of the battery cell; An image segmentation module configured to segment the pole piece image and extract the pole piece area to obtain a mask image; A feature extraction module configured to extract pixel features from the mask image; A pixel recognition module configured to identify the extracted pixel features to distinguish normal pixel points and abnormal pixel points; A foil leakage determination module configured to perform clustering analysis on the abnormal pixel points to determine the foil leakage area.

7. The coating leakage foil detection device in the production process of the battery cell according to claim 6, characterized in that, The segmenting the pole piece image includes: performing image segmentation using a pre-trained SAM model; the identifying the extracted pixel features includes: performing image recognition using a pre-trained GMM model.

8. An electronic device, characterized in that, Including a processor and a storage medium; The storage medium is used to store instructions; The processor is used to operate according to the instructions to execute the steps of the method according to any one of claims 1-5.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, the steps of the method according to any one of claims 1-5 are implemented.

10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, the steps of the method according to any one of claims 1-5 are implemented.

Images