System and method for automatically detecting and sorting lithium battery pole pieces

Abstract

The invention discloses an automatic detection and sorting system and method for lithium battery pole pieces, and the method comprises the steps: carrying out the state feature extraction of first and second state image information through employing an image analysis and extraction technology based on machine vision; performing homography transformation on the pole piece standard state image according to the first pole piece state feature and the second pole piece state feature, and performing state feature extraction on the commutation pole piece standard state image obtained through transformation; this intelligently generates the sorting information as a function of a priori modulation joint contrast perception representation between the second pole piece state feature and the pole piece standard state feature. Therefore, different types of pole piece defects can be automatically learned and identified so as to better adapt to various different defect forms. And meanwhile, multi-level and multi-scale characteristics can be extracted from the pole piece image, and tiny defect areas showing significant differences in comparison are highlighted, so that the accuracy and reliability of detection are improved.

Description

Technical Field

[0001] The present application relates to the field of lithium batteries, and more specifically, to a lithium battery pole piece automatic detection and sorting system and method. Background Art

[0002] Lithium battery pole pieces are core components of lithium batteries in electric vehicles and mobile devices, and their quality is directly related to battery performance and safety. Quality problems such as burrs generated during the die-cutting process will reduce the dimensional accuracy and surface quality of the pole pieces, and may puncture the diaphragm during battery assembly, causing safety risks. However, traditional lithium battery pole piece quality inspection mainly relies on manual sampling, which is inefficient and prone to missed inspections and false inspections.

[0003] In this regard, patent CN114384079B proposes a lithium battery pole piece burr detection and sorting system and method, which includes die-cutting, automatic detection, collection (qualified and burred pole pieces) and manual re-inspection units. The automatic detection unit distinguishes whether the pole piece has burrs based on the second image information and combines the first image information with the standard image information, and sends different types of pole pieces to the corresponding collection unit. In addition, the system is equipped with a reversing and transmission unit to optimize the process and reduce human errors.

[0004] In this patent, the second image information and the first image information are compared with the standard image through a specific algorithm to generate sorting information. However, the specific algorithm is usually based on predefined rules or parameters, and may not provide sufficient adaptability and generalization capabilities for different types of pole piece defects (such as burrs, cracks, etc.), which may lead to poor detection results when encountering new or unknown types of defects. In addition, simply comparing the second image information and the first image information with the standard image may not fully explore and utilize the deep features in the image. For example, for defect detection such as burrs on the edge of the pole piece, if you simply compare the images, you may ignore more detailed feature information such as the subtle shape and length of the burr, which is very important for accurately judging the quality of the pole piece.

[0005] Therefore, an optimized automatic detection and sorting solution for lithium battery electrodes is desired. Summary of the invention

[0006] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a lithium battery pole piece automatic detection and sorting system and method, which uses machine vision-based image analysis and extraction technology to extract state features of the first and second state image information, then performs homography transformation on the pole piece standard state image according to the first pole piece state feature and the second pole piece state feature, and then extracts state features of the transformed commutation pole piece standard state image, thereby intelligently generating the sorting information based on the prior modulation joint contrast perception representation between the second pole piece state feature and the pole piece standard state feature. The present application can automatically learn and identify different types of pole piece defects (such as burrs, cracks, etc.), rather than relying solely on predefined rules or parameters, which enables the system to better adapt to various defect forms. At the same time, multi-level and multi-scale features can be extracted from the pole piece image, and those tiny defect areas that show significant differences in comparison can be highlighted, thereby improving the accuracy and reliability of detection.

[0007] According to one aspect of the present application, a method for automatic detection and sorting of lithium battery pole pieces is provided, which includes: die-cutting a raw material sheet to obtain a plurality of die-cut lithium battery pole pieces; and automatically detecting and sorting each of the plurality of die-cut lithium battery pole pieces.

[0008] According to another aspect of the present application, a lithium battery pole piece automatic detection and sorting system is provided, which includes: a die-cutting module for die-cutting a raw material sheet to obtain a plurality of die-cut lithium battery pole pieces; an automatic detection and sorting module for automatically detecting and sorting each of the plurality of die-cut lithium battery pole pieces.

[0009] Compared with the prior art, the present application provides a lithium battery pole piece automatic detection and sorting system and method, which uses machine vision-based image analysis and extraction technology to extract state features of the first and second state image information, then performs homography transformation on the pole piece standard state image according to the first pole piece state feature and the second pole piece state feature, and then extracts state features from the transformed commutation pole piece standard state image, thereby intelligently generating the sorting information based on the prior modulation joint contrast perception representation between the second pole piece state feature and the pole piece standard state feature. The present application can automatically learn and identify different types of pole piece defects (such as burrs, cracks, etc.), rather than relying solely on predefined rules or parameters, which enables the system to better adapt to various defect forms. At the same time, multi-level and multi-scale features can be extracted from the pole piece image, and those tiny defect areas that show significant differences in comparison can be highlighted, thereby improving the accuracy and reliability of detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0011] Figure 1 The present invention is a flowchart of a method for automatically detecting and sorting lithium battery pole pieces according to an embodiment of the present application.

[0012] Figure 2 Schematic diagram of data flow of the method for automatic detection and sorting of lithium battery pole pieces according to an embodiment of the present application.

[0013] Figure 3 This is a flowchart of sub-step S2 of the method for automatically detecting and sorting lithium battery pole pieces according to an embodiment of the present application.

[0014] Figure 4 This is a flowchart of sub-step S24 of the method for automatically detecting and sorting lithium battery pole pieces according to an embodiment of the present application.

[0015] Figure 5 It is a block diagram of a lithium battery electrode automatic detection and sorting system according to an embodiment of the present application. DETAILED DESCRIPTION

[0016] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0017] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0018] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0019] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0020] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0021] In the technical solution of the present application, a method for automatic detection and sorting of lithium battery pole pieces is proposed. Figure 1 The present invention is a flowchart of a method for automatically detecting and sorting lithium battery pole pieces according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of the method for automatic detection and sorting of lithium battery pole pieces according to an embodiment of the present application. Figure 1 and Figure 2 As shown, the method for automatic detection and sorting of lithium battery pole pieces according to the embodiment of the present application includes the steps of: S1, die-cutting the raw material sheet to obtain a plurality of die-cut lithium battery pole pieces; S2, automatically detecting and sorting each of the plurality of die-cut lithium battery pole pieces.

[0022] In particular, the S1, die-cuts the raw material sheet to obtain a plurality of die-cut lithium battery pole pieces. In one example, before the formal large-scale production begins, the pre-cutting mode is usually adopted. At this time, the die-cutting unit will perform a single die-cut on the sheet, with the purpose of obtaining a small number of samples for preliminary testing and equipment debugging. In this way, the staff can check the appearance, size and burr of the small sample under a microscope, and adjust the parameters of the die-cutting equipment according to the test results until the quality requirements are met. Once the sample in the pre-cutting mode meets the standard, the die-cutting unit will switch to the working mode and start to continuously die-cut a large number of pole pieces from the sheet. In this process, each pole piece will be transported one by one to the subsequent processing steps. Whether it is pre-cutting or working mode, the task of the die-cutting unit is to accurately cut the original sheet into individual pole pieces according to the predetermined shape and size. This process needs to ensure that the cutting edge is smooth and there are no excessive burrs, because too large burrs may puncture the diaphragm during the subsequent battery assembly, causing safety problems such as short circuit or even explosion. The die-cut pole piece will be immediately conveyed to the commutation unit. Depending on the operating mode (pre-cutting or working), the reversing unit will decide whether to send the pole piece to the manual inspection unit or the automatic inspection unit. For the pre-cutting mode, the pole piece will be sent directly to the manual inspection unit; in the working mode, it will enter the automatic inspection unit for further quality inspection. When the die-cutting unit is in working mode, it will be located in a straight line with the reversing unit and the automatic inspection unit, that is, the first movement direction. This means that the pole piece is transported to each inspection point in sequence along this path. If it is in the pre-cutting mode, the reversing unit will turn the pole piece to another vertical direction-the second movement direction, and then send it to the manual inspection unit for detailed visual inspection. This design not only improves the flexibility of the process, but also provides conditions for subsequent automated inspection.

[0023] In particular, the step S2 automatically detects and sorts each of the plurality of die-cut lithium battery pole pieces. Specifically, in a specific example of the present application, Figure 3 As shown, the S2 includes: S21, collecting the first state image information of the die-cut lithium battery pole piece through the first camera; S22, selecting the pole piece standard state image with the die-cutting quality marked as qualified from the database; S23, generating commutation information based on the first state image information and the pole piece standard state image; S24, after reversing the die-cut lithium battery pole piece based on the commutation information, collecting the second state image information of the die-cut lithium battery pole piece through the first camera; S25, obtaining the sorting information of the die-cut lithium battery pole piece based on the first state image information and the second state image information and in combination with the pole piece standard state image; S26, based on the sorting information, transporting the die-cut lithium battery pole piece to the corresponding area.

[0024] Specifically, the S21, collects the first state image information of the die-cut lithium battery pole piece through the first camera. In one example, the first camera is usually located above the carrying surface and keeps a certain distance from the carrying surface to ensure that the shooting angle and range are suitable for obtaining high-quality image information. The camera is equipped with a positive light source to provide uniform front lighting to ensure that the image is clearly visible. In addition, the backlight source is arranged below the carrying surface and coaxially with the positive light source to enhance the contrast, which is particularly useful when detecting edge details such as burrs. This light source configuration helps to improve image quality and makes subsequent feature extraction more accurate. When a die-cut lithium battery pole piece is transferred to the automatic detection unit, the motion module places it on the carrying surface. At this time, the pole piece on the carrying surface is in the initial position and is ready to accept the first image acquisition. Once the pole piece stabilizes, the control system sends a signal to the first camera to trigger it to start shooting. This trigger mechanism ensures that each shot is taken when the pole piece is completely still, thereby avoiding blurring caused by motion. The first camera quickly captures the surface features of the pole piece, including color, texture, and any possible defects such as burrs or cracks. These images are called "first state image information", which records the state of the polar surface before any transformation. After shooting, the first state image information will be saved and transmitted to the image processing unit in the control system in real time. The data transmission speed here needs to be fast enough to ensure that the subsequent steps will not be affected by data delays.

[0025] Specifically, the S22 selects a standard state image of a pole piece with a die-cutting quality mark as qualified from the database. It should be understood that in the production process of lithium battery pole pieces, it is crucial to ensure the quality of each pole piece. In order to achieve this goal, a high-precision first camera is used in the automatic detection system to collect the first state image information of the lithium battery pole piece after die-cutting. However, only high-quality image acquisition is not enough to ensure that the quality inspection of the pole piece reaches the highest standard. It should be understood that using a standard image with a die-cutting quality mark as qualified as a reference can ensure that every pole piece shipped from the factory meets the established quality standards. This not only helps to maintain the quality consistency of the product, but also provides a reliable foundation for subsequent battery assembly and reduces potential safety hazards. The establishment and improvement of the standard image library is a continuous process that reflects the progress and development of the production process. Through the selection and use of standard images, the system can capture the latest production technology and process improvements in a timely manner, and feedback to the production department for optimization and adjustment, forming a virtuous circle and promoting the continuous improvement of the entire production process. Therefore, after image acquisition, a reliable comparison benchmark is needed to determine whether the pole piece meets the quality requirements. In the technical solution of the present application, a standard state image of a pole piece with a die-cutting quality mark as qualified is selected from the database. Each standard image will be annotated in detail, including but not limited to the size, shape, color of the pole piece and any specific manufacturing parameters (such as die-cutting process parameters). In addition, the images will be classified and stored according to their source, production batch, production equipment and other factors, so that the most suitable reference image can be quickly located during subsequent retrieval. This detailed annotation and classification not only helps to improve retrieval efficiency, but also enables the system to select the most suitable standard image as a reference in different scenarios.

[0026] Specifically, the S23 generates commutation information based on the first state image information and the pole piece standard state image. Specifically, the first state image information is compared with the pole piece standard state image for quality analysis. If potential problems are found on one side or in certain areas of the pole piece (such as possible burrs or other quality problems), the control system will decide whether it is necessary to change the direction of the pole piece surface for further inspection. If it is determined that commutation is required, the control component will generate corresponding commutation information. The commutation information contains data indicating how the commutation component moves, such as the angle of the turn or the specific direction of rotation. When the commutation component receives the commutation information, the current movement is commutated so that the pole piece orientation is turned to ensure that all edges of the pole piece can be fully inspected, especially the die-cut edges and the slitting edges, to ensure that no possible defects are missed.

[0027] Specifically, in S24, after the die-cut lithium battery pole piece is reversed based on the reversal information, the second state image information of the die-cut lithium battery pole piece is collected by the first camera. Specifically, in a specific example of the present application, Figure 4 As shown, the S24 includes: S241, performing pole piece state feature analysis on the first state image information and the second state image information to obtain a first pole piece state feature and a second pole piece state feature; S242, based on the first pole piece state feature and the second pole piece state feature, performing a homography transformation on the pole piece standard state image to obtain a commutation pole piece standard state image; S243, performing the pole piece state feature analysis on the commutation pole piece standard state image to obtain a pole piece standard state feature; S244, performing a priori modulation of joint feature perception on the second pole piece state feature and the pole piece standard state feature to obtain a pole piece state joint contrast perception feature, and obtaining the sorting information based on the pole piece state joint contrast perception feature.

[0028] More specifically, the S241 performs a pole piece state feature analysis on the first state image information and the second state image information to obtain a first pole piece state feature and a second pole piece state feature. Considering that both the first state image information and the second state image information contain some key representative feature information of the surface features of the lithium battery pole piece, such as the shape, edge, surface texture, tiny burrs, cracks or other minor defects of the pole piece. Therefore, in the technical solution of the present application, the first state image information and the second state image information are processed by a pole piece state feature extractor based on the FCN network to obtain a first pole piece state feature map and a second pole piece state feature map. It can be understood that when processing image data, the FCN network can utilize the parameter sharing characteristics of the convolutional layer to efficiently process a large amount of image pixel data. For images of lithium battery pole pieces that may contain a large number of tiny details (such as potential burrs), the FCN network can quickly extract features from the entire image without losing important detail information, reducing the waste of computing resources and improving the efficiency of feature extraction, so as to extract features related to the pole piece status, such as whether there are burrs on the edge of the pole piece and whether the surface is flat.

[0029] More specifically, in S242, based on the first pole piece state feature and the second pole piece state feature, the pole piece standard state image is subjected to homography transformation to obtain the commutation pole piece standard state image. That is, in the technical solution of the present application, first, based on the first pole piece state feature map and the second pole piece state feature map, a homography matrix between the first state image information and the second state image information is constructed; considering that the first state image information and the second state image information correspond to the image presentation of the lithium battery pole piece in different states, the pole piece has undergone commutation operation and other processing, and its position, angle and relative geometric relationship between the parts in the image have changed. Therefore, in order to better reflect the change relationship that occurs in the morning of this plane, in the technical solution of the present application, based on the first pole piece state feature map and the second pole piece state feature map, a homography matrix between the first state image information and the second state image information is constructed. It is worth mentioning that the homography matrix is ​​a 3×3 matrix used to describe the perspective transformation relationship between two two-dimensional images on a plane. In computer vision, when a planar object (such as a lithium battery pole piece) is photographed into two images at different viewing angles, positions, or after transformations such as rotation and scaling, the homography matrix can map the coordinates of a point in one image to the coordinates of the corresponding point in another image through a linear transformation, which can well describe the two-dimensional geometric transformation relationship on this plane. It can associate the coordinates of the pixel points in the first state image with the coordinates of the corresponding pixel points in the second state image, and accurately depict the changes in the image level when the pole piece changes from one state to another. Further, considering that the first state image information and the second state image information are actual detection images of the lithium battery pole piece in different states, a geometric transformation relationship is established between them through the homography matrix. In order to enable the pole piece standard state image to be effectively compared with the actual detected pole piece state (especially the second state, because it can more comprehensively reflect the pole piece situation after the commutation operation), in the technical solution of the present application, based on the homography matrix, the pole piece standard state image is homographically transformed to obtain the commutation pole piece standard state image. This ensures that the standard state image and the actual second state pole piece image are consistent in spatial geometric relationship, and the standard state image can be adjusted to the same state as the pole piece after commutation to avoid inaccurate feature comparison due to geometric differences. For example, if the homography transformation is not performed, the pole piece edges and key feature areas in the standard state image and the actual second state pole piece image cannot match in spatial position and shape, which will seriously affect the subsequent judgment of whether the pole piece has burrs and other quality problems.

[0030] In one example, the process of constructing a homography matrix involves multiple aspects. The system searches for corresponding key points or feature areas between the first pole piece state feature map and the second pole piece state feature map and establishes a mapping relationship between them. Based on the key point pairs found, an algorithm (such as direct linear transformation DLT) is used to estimate the parameters of the homography matrix. This process usually involves minimizing an error function to ensure that the estimated homography matrix can most accurately describe the transformation relationship between the two images. To further improve the accuracy of the homography matrix, the system may use iterative optimization methods such as RANSAC (random sampling consensus) to remove outliers and recalculate the best parameter combination. Once the homography matrix is ​​obtained, it can be applied to the pole piece standard state image to obtain the commutation pole piece standard state image. The main purpose of this step is to adjust the angle and position of the standard image so that it is geometrically consistent with the actual collected pole piece image. Specifically, the homography matrix is ​​used to transform the coordinates of each pixel point in the standard image into coordinates in the new coordinate system. This process can be achieved by matrix multiplication, that is, multiplying the homogeneous coordinate form of the original coordinate by the homography matrix to obtain a new homogeneous coordinate, and then converting it back to the normal coordinate form. Since coordinate transformation may cause some pixels to no longer be located at integer coordinate positions, image resampling is required. Common resampling methods include nearest neighbor interpolation, bilinear interpolation, and bicubic interpolation. Choosing a suitable interpolation method can minimize the distortion caused by coordinate transformation while ensuring image quality. Considering that the pole piece may be partially occluded or exceed the window range at different viewing angles, the system also needs to perform appropriate boundary processing on the transformed image. For example, for areas beyond the boundaries of the original image, you can choose to fill the background color or crop it; for the occluded part, you can make reasonable inferences based on the colors of the surrounding pixels to fill it.

[0031] More specifically, the S243 performs the pole piece state feature analysis on the standard state image of the commutating pole piece to obtain the pole piece standard state feature. In the technical solution of the present application, the standard state image of the commutating pole piece is passed through the pole piece state feature extractor based on the FCN network to obtain the pole piece standard state feature map as the pole piece standard state feature. It should be understood that although the standard state image of the commutating pole piece is geometrically consistent with the second state image of the actual pole piece after the homography transformation, the original image data still contains a large amount of pixel information, some of which may be redundant or interfering for judging the quality of the pole piece. Therefore, in order to focus on the key feature information in the standard state image of the commutating pole piece, so as to make a more accurate comparison and analysis with the second pole piece state feature map later, in the technical solution of the present application, the standard state image of the commutating pole piece is passed through the pole piece state feature extractor based on the FCN network to utilize the powerful feature learning ability of the FCN network to automatically identify and extract the features that are critical to the pole piece state, such as edge, texture and burr information, to obtain the pole piece standard state feature map. In this way, the second pole piece state characteristic diagram can be compared with the standard characteristic diagram, so that the difference in characteristics between the actual pole piece and the standard pole piece can be clearly seen.

[0032] More specifically, in S244, the second pole piece state feature and the pole piece standard state feature are subjected to a priori modulation of joint feature perception to obtain a pole piece state joint comparison perception feature, and the sorting information is obtained based on the pole piece state joint comparison perception feature. That is, in the technical solution of the present application, first, the second pole piece state feature diagram and the pole piece standard state feature diagram are subjected to a priori modulation of joint feature perception to obtain a pole piece state joint comparison perception feature diagram as the pole piece state joint comparison perception feature; considering that the second pole piece state feature diagram contains the key surface state feature information of the actually detected lithium battery pole piece in a specific state (the second state after commutation and other operations), and the pole piece standard state feature diagram represents the standard key state feature of a qualified pole piece. In order to comprehensively and accurately evaluate the difference between the actual pole piece and the standard pole piece, in the technical solution of the present application, the second pole piece state feature diagram and the pole piece standard state feature diagram are subjected to a priori modulation of joint feature perception to obtain a pole piece state joint comparison perception feature diagram. In particular, simple feature difference calculation may not be able to fully consider the complex relationship between features and the importance of different features for pole piece quality judgment, while the joint feature-aware prior modulation can utilize prior knowledge, such as the standards for pole piece quality and the contribution of different features to quality detection, to identify and highlight these significant features, so as to more reasonably compare the two feature maps, thereby highlighting those feature differences that are closely related to quality judgment.

[0033] Specifically, first, the second pole piece state characteristic diagram and the pole piece standard state characteristic diagram are subjected to characteristic shape reshaping to obtain a second pole piece state characteristic shape reshaping matrix and a pole piece standard state characteristic shape reshaping matrix. In a specific example, each local characteristic matrix along the channel dimension of the second pole piece state characteristic diagram and the pole piece standard state characteristic diagram is expanded into a characteristic vector to obtain the second pole piece state characteristic shape reshaping matrix (HW×C) and the pole piece standard state characteristic shape reshaping matrix. More specifically, the second pole piece state characteristic diagram and the pole piece standard state characteristic diagram are subjected to characteristic shape reshaping using the following characteristic shape reshaping formula to obtain a second pole piece state characteristic shape reshaping matrix and a pole piece standard state characteristic shape reshaping matrix; wherein, the characteristic shape reshaping formula is: ;in, is the characteristic diagram of the second pole piece state, is the characteristic diagram of the pole piece standard state, For the reshape operation, is the second pole piece state characteristic shape reshaping matrix, is the pole piece standard state characteristic shape reshaping matrix.

[0034] Next, the second pole piece state characteristic shape reshaping matrix and the pole piece standard state characteristic shape reshaping matrix are subjected to a priori guided feature modulation interaction to obtain the second pole piece state interaction encoding feature matrix and the pole piece standard state interaction encoding feature matrix. Specifically, first, the second pole piece state characteristic shape reshaping matrix and the pole piece standard state characteristic shape reshaping matrix are subjected to a priori guided modulation to obtain the prior modulated second pole piece state characteristic shape reshaping matrix and the prior modulated pole piece standard state characteristic shape reshaping matrix; wherein the prior information is information pre-set according to domain knowledge, historical data or other sources to guide the model to better understand the input data. In this process, the prior information is used as a gating signal to determine which features should be amplified and which should be weakened. Then, the prior modulated second pole piece state characteristic shape reshaping matrix and the prior modulated pole piece standard state characteristic shape reshaping matrix are subjected to a converter-like attention interaction to obtain the second pole piece state interaction encoding feature matrix and the pole piece standard state interaction encoding feature matrix. The core of the converter structure lies in the self-attention mechanism, which enables the model to pay attention to any position in the entire sequence when processing sequence data. In the technical solution of the present application, the information of each position in the prior modulated second pole piece state characteristic shape reshaping matrix can interact globally and finely with the information of each position in the prior modulated pole piece standard state characteristic shape reshaping matrix to generate the second pole piece state interaction coding feature matrix and the pole piece standard state interaction coding feature matrix containing fine-grained unidirectional interactive global information. More specifically, the second pole piece state characteristic shape reshaping matrix and the pole piece standard state characteristic shape reshaping matrix are subjected to a priori guided feature modulation interaction with the following feature modulation interaction formula to obtain the second pole piece state interaction coding feature matrix and the pole piece standard state interaction coding feature matrix; wherein the feature modulation interaction formula is: ;in, and represents the prior-guided modulation learnable memory parameter matrix, for The transposed matrix of is matrix multiplication, represents the normalization function, is the prior modulated second pole piece state characteristic shape reshaping matrix, is the prior modulated pole piece standard state characteristic shape reshaping matrix, yes The transposed matrix of yes The transposed matrix of and They are and The scale is the matrix width multiplied by the matrix height, is the normalization function, is the second pole piece state interaction encoding feature matrix, It is the pole piece standard state interaction encoding feature matrix.

[0035] Then, the second pole piece state interactive coding feature matrix and the pole piece standard state interactive coding feature matrix are reshaped and weighted fused to obtain the pole piece state joint contrast perception feature map. Specifically, first, the second pole piece state interactive coding feature matrix and the pole piece standard state interactive coding feature matrix are reshaped to obtain the second pole piece state interactive coding feature map and the pole piece standard state interactive coding feature map. That is, after completing the attention interaction coding, in order to restore the feature matrix to a spatial structure that matches the original feature map, so as to facilitate subsequent processing or combination with other feature maps, the feature shape reshaping will be performed again. In one example, the spatial resolution of the feature map can be increased by transposed convolution, or an interpolation method can be used to smooth the feature transition. Then, the weighted sum of the second pole piece state interactive coding feature map and the pole piece standard state interactive coding feature map is calculated to obtain the pole piece state joint contrast perception feature map. That is, the second pole piece state interactive coding feature map and the pole piece standard state interactive coding feature map are integrated by weighted difference. More specifically, the second pole piece state interactive coding feature matrix and the pole piece standard state interactive coding feature matrix are reshaped and weighted fused by the following reshape reshaping and weighted fusion formula to obtain the pole piece state joint contrast perception feature map; wherein the reshape reshaping and weighted fusion formula is: ;in, is the interactive coding feature map of the second pole piece state, is the standard state interactive coding feature diagram of the pole piece, and is the weighted hyperparameter, It is the joint contrast perception feature map of the pole piece state.

[0036] Furthermore, based on the electrode state joint comparison perception feature map, the sorting information is generated, wherein the sorting information is whether the die-cut lithium battery electrode is a qualified lithium battery electrode. In the technical solution of the present application, the electrode state joint comparison perception feature map is passed through a classifier-based lithium battery electrode automatic sorter to generate the sorting information. That is, the electrode state joint comparison perception feature obtained by a priori joint comparison of the second electrode state feature map and the electrode standard state feature map is used for classification processing, so as to accurately generate sorting information about lithium battery electrodes, and clearly indicate whether each electrode belongs to the category of qualified or unqualified (such as the presence of burrs and other quality defects). This allows the electrodes of different quality states to be transported to the corresponding areas according to the sorting information in the subsequent production process, such as sending qualified electrodes to subsequent assembly links, and sending unqualified electrodes to rework or scrapping.

[0037] In a preferred example, when the second pole piece state feature map and the pole piece standard state feature map respectively represent the image semantic features of the second state image information and the image semantic features of the commutated pole piece standard state image based on the state feature homography transformation response, when performing feature significant joint contrast perception based on prior guidance, the prior guidance misalignment of the image semantic feature pattern caused by the semantic difference of the source image will also cause the significant joint contrast perception distribution richness of the pole piece state joint contrast perception feature map, resulting in repetition of unstructured probability mapping, and thus it is necessary to improve the feature space-class target probability response effect of the pole piece state joint contrast perception feature map.

[0038] Based on this, the applicant of the present application optimizes the pole piece state joint contrast perception feature map when passing the pole piece state joint contrast perception feature map through a classifier-based lithium battery pole piece automatic sorter to generate the sorting information, including: calculating the sum of the absolute values ​​of all feature values ​​of the pole piece state joint contrast perception feature map , and the square root of the sum of the squares of all eigenvalues ​​of the joint contrast perception feature map of the pole piece state , and Divide by To obtain the joint contrast perception probability value of the pole piece state ,Right now: ;in, The first one represents the joint contrast perception feature map of the electrode state feature values.

[0039] The electrode state is jointly compared with the perception feature map Each eigenvalue in is normalized to the maximum value to obtain the joint contrast perception probability feature map of the pole piece state : ;in, represents the maximum value in the joint contrast perception feature map of the pole piece state, The first one represents the joint contrast perception probability feature map of the pole piece state feature values.

[0040] After the pole piece state joint contrast perception probability feature map is subtracted from the pole piece state joint contrast perception probability value, it is further multiplied by the reciprocal of the pole piece state joint contrast perception probability value to obtain the pole piece state joint contrast perception domain gradient feature map ,in, Indicates point reduction, Represents dot product.

[0041] After subtracting the pole piece state joint contrast perception probability feature map from the unit feature map, the pole piece state joint contrast perception domain gradient feature map is divided by the point to obtain the pole piece state joint contrast perception differential feature map, that is: ;in, represents the unit eigenvector, represents the bit-by-bit inverse of the pole piece state joint contrast perception domain gradient feature map, that is, calculating the inverse of each eigenvalue of the pole piece state joint contrast perception domain gradient feature map, Represents the joint contrast perception differential feature map of the pole piece state.

[0042] Calculate the power function of the pole piece state joint contrast perception domain gradient feature map with the pole piece state joint contrast perception probability value as the exponent , and the exponential function with the natural constant as the base of the perception differential characteristic diagram is jointly compared with the pole piece state The weighted sum is performed to obtain the optimized pole piece state joint contrast perception feature map, namely: ;in, and represents the weighted hyperparameter, Indicates point addition, Represents the optimized pole piece state joint contrast perception feature map.

[0043] That is, the mutually exclusive generalized representation behavior units of the joint contrast perception feature map of the pole piece state are constructed through the discrete differential of the probability density domain gradient corresponding to the joint contrast perception feature map of the pole piece state, so as to enhance the static response characteristics of the short-range correlation mesoscopic information configuration of the joint contrast perception feature map of the pole piece state to the generative probabilistic generalized representation behavior based on the different behavior unit organization spaces under the non-uniform topological architecture of the joint contrast perception feature map of the pole piece state, thereby ensuring the steady-state convergence of the optimization process between the generated target and the extracted features in the feature space-generated probability mapping, so as to improve the accuracy of the classification results obtained by the automatic lithium battery pole sorter based on the classifier input of the joint contrast perception feature map of the pole piece state.

[0044] Specifically, the S25 and S26 obtain the sorting information of the die-cut lithium battery pole piece based on the first state image information and the second state image information and in combination with the pole piece standard state image; and based on the sorting information, transport the die-cut lithium battery pole piece to the corresponding area.

[0045] In particular, in one example, different pole pieces are transported to the corresponding collection units according to the sorting information - qualified pole pieces are sent to the second collection unit, and burred pole pieces are sent to the first collection unit for further manual re-inspection. If there are too many burred pole pieces in the first collection unit at the same time, it means that there may be an abnormality in the die-cutting unit. At this time, the system will automatically generate a shutdown message, stop the die-cutting and detection process, prevent more defective products from being produced, and remind the operator to check and adjust the equipment parameters in time.

[0046] In summary, according to the embodiment of the present application, the automatic detection and sorting method of lithium battery pole pieces is explained, which uses machine vision-based image analysis and extraction technology to extract state features of the first and second state image information, and then performs homography transformation on the pole piece standard state image according to the first pole piece state feature and the second pole piece state feature, and then extracts state features of the transformed commutation pole piece standard state image, thereby intelligently generating the sorting information based on the prior modulation joint contrast perception representation between the second pole piece state feature and the pole piece standard state feature. The present application can automatically learn and identify different types of pole piece defects (such as burrs, cracks, etc.), rather than relying solely on predefined rules or parameters, which enables the system to better adapt to various defect forms. At the same time, multi-level and multi-scale features can be extracted from the pole piece image, and those tiny defect areas that show significant differences in comparison can be highlighted, thereby improving the accuracy and reliability of detection.

[0047] Furthermore, a lithium battery pole piece automatic detection and sorting system is also provided.

[0048] Figure 5FIG. 1 is a block diagram of a lithium battery electrode automatic detection and sorting system according to an embodiment of the present application. Figure 5 As shown, the automatic detection and sorting 300 of lithium battery pole pieces according to the embodiment of the present application includes: a die-cutting module 310, which is used to die-cut the raw material sheet to obtain a plurality of die-cut lithium battery pole pieces; an automatic detection and sorting module 320, which is used to automatically detect and sort each of the plurality of die-cut lithium battery pole pieces.

[0049] As described above, the lithium battery pole piece automatic detection and sorting system 300 according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with a lithium battery pole piece automatic detection and sorting algorithm. In one possible implementation, the lithium battery pole piece automatic detection and sorting system 300 according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the lithium battery pole piece automatic detection and sorting system 300 can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the lithium battery pole piece automatic detection and sorting system 300 can also be one of the many hardware modules of the wireless terminal.

[0050] Alternatively, in another example, the lithium battery pole piece automatic detection and sorting system 300 and the wireless terminal may also be separate devices, and the lithium battery pole piece automatic detection and sorting system 300 may be connected to the wireless terminal via a wired and / or wireless network, and transmit interactive information in accordance with an agreed data format.

[0051] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method for automatically detecting and sorting lithium battery pole pieces, comprising: Die-cutting the raw material sheet to obtain a plurality of die-cut lithium battery pole pieces; Automatically detect and sort each of the plurality of die-cut lithium battery pole pieces, characterized in that the automatic detection and sorting of each of the plurality of die-cut lithium battery pole pieces includes: collecting first state image information of the die-cut lithium battery pole piece through a first camera; selecting a pole piece standard state image with a die-cutting quality mark as qualified from a database; generating commutation information based on the first state image information and the pole piece standard state image; after reversing the die-cut lithium battery pole piece based on the commutation information, collecting second state image information of the die-cut lithium battery pole piece through the first camera; obtaining sorting information of the die-cut lithium battery pole piece based on the first state image information and the second state image information and in combination with the pole piece standard state image; and sorting the die-cut lithium battery pole piece based on the sorting information. The sheet is transported to the corresponding area; wherein, based on the first state image information and the second state image information and in combination with the electrode standard state image, the sorting information of the lithium battery electrode after die-cutting is obtained, including: performing a electrode state feature analysis on the first state image information and the second state image information to obtain a first electrode state feature and a second electrode state feature; based on the first pole piece state feature and the second pole piece state feature, performing a homography transformation on the pole piece standard state image to obtain a commutation pole piece standard state image; performing the pole piece state feature analysis on the commutation pole piece standard state image to obtain a pole piece standard state feature; performing a priori modulation of joint feature perception on the second pole piece state feature and the pole piece standard state feature to obtain a pole piece state joint contrast perception feature, and obtaining the sorting information based on the pole piece state joint contrast perception feature.

2. The method for automatic detection and sorting of lithium battery pole pieces according to claim 1, characterized in that: Performing pole piece state feature analysis on the first state image information and the second state image information to obtain a first pole piece state feature and a second pole piece state feature, including: passing the first state image information and the second state image information through a pole piece state feature extractor based on an FCN network to obtain a first pole piece state feature diagram as the first pole piece state feature and a second pole piece state feature diagram as the second pole piece state feature.

3. The method for automatic detection and sorting of lithium battery pole pieces according to claim 2, characterized in that: Based on the first pole piece state characteristics and the second pole piece state characteristics, the pole piece standard state image is subjected to a homography transformation to obtain a commutation pole piece standard state image, including: based on the first pole piece state characteristic diagram and the second pole piece state characteristic diagram, constructing a homography matrix between the first state image information and the second state image information; based on the homography matrix, performing a homography transformation on the pole piece standard state image to obtain the commutation pole piece standard state image.

4. The method for automatic detection and sorting of lithium battery pole pieces according to claim 3, characterized in that: The pole piece state feature analysis is performed on the commutator pole piece standard state image to obtain the pole piece standard state feature, including: passing the commutator pole piece standard state image through the pole piece state feature extractor based on the FCN network to obtain a pole piece standard state feature diagram as the pole piece standard state feature.

5. The method for automatic detection and sorting of lithium battery pole pieces according to claim 4, characterized in that: The second pole piece state feature and the pole piece standard state feature are subjected to a priori modulation of joint feature perception to obtain a pole piece state joint comparison perception feature, and the sorting information is obtained based on the pole piece state joint comparison perception feature, including: the second pole piece state feature diagram and the pole piece standard state feature diagram are subjected to a priori modulation of joint feature perception to obtain a pole piece state joint comparison perception feature diagram as the pole piece state joint comparison perception feature; based on the pole piece state joint comparison perception feature diagram, the sorting information is generated, wherein the sorting information is whether the die-cut lithium battery pole piece is a qualified lithium battery pole piece.

6. The method for automatic detection and sorting of lithium battery pole pieces according to claim 5, characterized in that: The second pole piece state characteristic diagram and the pole piece standard state characteristic diagram are subjected to a priori modulation of joint feature perception to obtain a pole piece state joint contrast perception characteristic diagram, including: performing feature shape reshaping on the second pole piece state characteristic diagram and the pole piece standard state characteristic diagram to obtain a second pole piece state feature shape reshaping matrix and a pole piece standard state feature shape reshaping matrix; performing a priori guided feature modulation interaction on the second pole piece state feature shape reshaping matrix and the pole piece standard state feature shape reshaping matrix to obtain a second pole piece state interactive coding feature matrix and a pole piece standard state interactive coding feature matrix; performing shape reshaping and weighted fusion on the second pole piece state interactive coding feature matrix and the pole piece standard state interactive coding feature matrix to obtain the pole piece state joint contrast perception characteristic diagram.

7. The method for automatic detection and sorting of lithium battery pole pieces according to claim 6, characterized in that: The second pole piece state characteristic shape reshaping matrix and the pole piece standard state characteristic shape reshaping matrix are subjected to a priori guided feature modulation interaction to obtain a second pole piece state interaction coding feature matrix and a pole piece standard state interaction coding feature matrix, including: performing a priori guided modulation on the second pole piece state characteristic shape reshaping matrix and the pole piece standard state characteristic shape reshaping matrix to obtain a priori modulated second pole piece state characteristic shape reshaping matrix and a priori modulated pole piece standard state characteristic shape reshaping matrix; performing converter-like attention interaction on the priori modulated second pole piece state characteristic shape reshaping matrix and the priori modulated pole piece standard state characteristic shape reshaping matrix to obtain the second pole piece state interaction coding feature matrix and the pole piece standard state interaction coding feature matrix.

8. The method for automatic detection and sorting of lithium battery pole pieces according to claim 7, characterized in that: The second pole piece state interactive coding feature matrix and the pole piece standard state interactive coding feature matrix are reshaped and weightedly fused to obtain the pole piece state joint contrast perception feature map, including: reshaping the feature shape of the second pole piece state interactive coding feature matrix and the pole piece standard state interactive coding feature matrix to obtain the second pole piece state interactive coding feature map and the pole piece standard state interactive coding feature map; calculating the weighted sum of the second pole piece state interactive coding feature map and the pole piece standard state interactive coding feature map to obtain the pole piece state joint contrast perception feature map.

9. The method for automatic detection and sorting of lithium battery pole pieces according to claim 8, characterized in that: Based on the pole piece state joint contrast perception feature map, the sorting information is generated, including: the pole piece state joint contrast perception feature map is passed through a classifier-based lithium battery pole piece automatic sorter to generate the sorting information.

10. A lithium battery pole piece automatic detection and sorting system, characterized in that: include: A die-cutting module, used for die-cutting the raw material sheet to obtain a plurality of die-cut lithium battery pole pieces; The automatic detection and sorting module is used to automatically detect and sort each of the die-cut lithium battery pole pieces among the plurality of die-cut lithium battery pole pieces.

Images