Lithium battery pole piece positioning method and system based on dynamic filter

Abstract

A lithium battery pole piece positioning method and system based on a dynamic filter, which obtains a lithium battery image to be processed and a reference lithium battery image; uses artificial intelligence technology based on deep learning to spatially enhance and dynamically filter the lithium battery image to be processed, thereby achieving accurate positioning and identification of the positive and negative pole pieces.

Description

Technical Field

[0001] The present application relates to the field of intelligent positioning technology, and more specifically, to a lithium battery pole piece positioning method and system based on a dynamic filter. Background Art

[0002] Since 2018, ternary lithium batteries have begun to become the absolute leader in the field of new energy vehicle power batteries with overwhelming advantages.

[0003] Correct positioning of the positive and negative electrodes can ensure the best performance of the battery. At present, multiple types of interference pose severe challenges to the accurate positioning of the positive and negative electrodes, including but not limited to: electrode bifurcation interference, interference from the tray being too close, electrode interference, tilt, electrode ear interference, and diaphragm interference. Therefore, a lithium battery electrode positioning solution is expected. Summary of the invention

[0004] In order to solve the above technical problems, this application is proposed. The embodiment of this application provides a lithium battery pole piece positioning method and system based on a dynamic filter, which obtains a lithium battery image to be processed and a reference lithium battery image; uses artificial intelligence technology based on deep learning to spatially enhance and dynamically filter the lithium battery image to be processed, thereby achieving accurate positioning and identification of the positive and negative pole pieces.

[0005] In a first aspect, a lithium battery pole piece positioning method based on a dynamic filter is provided, which comprises:

[0006] Acquire a lithium battery image to be processed and a reference lithium battery image;

[0007] Passing the to-be-processed lithium battery image through a first convolutional neural network model as a feature extractor to obtain a lithium battery detection feature map;

[0008] Passing the reference lithium battery image through a second convolutional neural network model as a feature extractor to obtain a lithium battery reference feature map;

[0009] Pass the lithium battery reference feature map through a spatial attention module to obtain a spatial attention weight matrix;

[0010] Calculate the point-by-point multiplication of each feature matrix along the channel dimension of the lithium battery detection feature map and the spatial attention weight matrix to obtain a reference enhanced lithium battery detection feature map;

[0011] Passing the reference enhanced lithium battery detection characteristic graph through a dynamic filtering module to obtain a filtered reference enhanced lithium battery detection characteristic graph; and

[0012] The filtered reference enhanced lithium battery detection characteristic image is passed through a decoder to obtain a filtered lithium battery image.

[0013] In the above-mentioned lithium battery pole piece positioning method based on dynamic filter, the lithium battery image to be processed is passed through the first convolutional neural network model as a feature extractor to obtain a lithium battery detection feature map, including: using each layer of the second convolutional neural network model as a feature extractor to perform convolution processing, feature matrix-based mean pooling processing and nonlinear activation processing on the input data in the forward transfer of the layer so that the last layer of the second convolutional neural network model as a feature extractor outputs the lithium battery detection feature map, wherein the input of the first layer of the second convolutional neural network model as a feature extractor is the lithium battery image to be processed.

[0014] In the above-mentioned lithium battery pole piece positioning method based on dynamic filter, the reference lithium battery image is passed through the second convolutional neural network model as a feature extractor to obtain a lithium battery reference feature map, including: using each layer of the second convolutional neural network model as a feature extractor to perform convolution processing, feature matrix-based mean pooling processing and nonlinear activation processing on the input data in the forward transfer of the layer so that the last layer of the second convolutional neural network model as a feature extractor outputs the lithium battery reference feature map, wherein the input of the first layer of the second convolutional neural network model as a feature extractor is the reference lithium battery image.

[0015] In the above-mentioned lithium battery pole piece positioning method based on dynamic filter, the lithium battery reference feature map is passed through the spatial attention module to obtain the spatial attention weight matrix, including: calculating the global mean of the lithium battery reference feature map along the channel dimension to obtain the lithium battery reference feature matrix; passing the lithium battery reference feature matrix through the convolution layer to obtain the spatial attention feature matrix; and, inputting the spatial attention feature matrix into the Softmax activation function to obtain the spatial attention weight feature matrix.

[0016] In the above-mentioned lithium battery pole piece positioning method based on dynamic filter, the dynamic filtering module includes at least one convolution layer.

[0017] In the above-mentioned lithium battery pole piece positioning method based on dynamic filter, it also includes a training step: training the first convolutional neural network model as a feature extractor, the second convolutional neural network model as a feature extractor, the spatial attention module, the dynamic filtering module and the decoder; wherein the training step includes: obtaining training data, the training data includes a training lithium battery image to be processed and a training reference lithium battery image, as well as a real filtered lithium battery image; passing the training lithium battery image to be processed through the first convolutional neural network model as a feature extractor to obtain a training lithium battery detection feature map; passing the training reference lithium battery image through the second convolutional neural network model as a feature extractor to obtain a training lithium battery reference feature map; passing the training lithium battery reference feature map through the spatial attention module to obtain a training spatial attention weight matrix; calculating each feature of the training lithium battery detection feature map along the channel dimension The matrix is ​​multiplied by the training spatial attention weight matrix at each position point to obtain a training reference enhanced lithium battery detection feature map; the training reference enhanced lithium battery detection feature map is subjected to a Fourier scale domain probability correction to obtain an optimized training reference enhanced lithium battery detection feature map; the optimized training reference enhanced lithium battery detection feature map is passed through the dynamic filtering module to obtain a training filtered reference enhanced lithium battery detection feature map; the training filtered reference enhanced lithium battery detection feature map is passed through the decoder to obtain a training filtered lithium battery image; the mean square error between the training filtered lithium battery image and the true filtered lithium battery image is calculated as a decoding loss function value; and, based on the decoding loss function value and through the direction propagation of gradient descent, the first convolutional neural network model as a feature extractor, the second convolutional neural network model as a feature extractor, the spatial attention module, the dynamic filtering module and the decoder are trained.

[0018] In the above-mentioned lithium battery pole piece positioning method based on dynamic filter, the training reference enhanced lithium battery detection characteristic map is subjected to Fourier scale domain probability correction to obtain the optimized training reference enhanced lithium battery detection characteristic map, including: the training reference enhanced lithium battery detection characteristic map is subjected to Fourier scale domain probability correction to obtain the optimized training reference enhanced lithium battery detection characteristic map according to the following optimization formula; wherein the optimization formula is:

[0019]

[0020] Wherein, F represents the reference enhanced lithium battery detection characteristic diagram, f i,j,k∈F is the eigenvalue of the (i, j, k)th position of the reference enhanced lithium battery detection feature map, W, H and C are the height, width and number of channels of the reference enhanced lithium battery detection feature map, respectively, and α, β and γ are hyperparameters for scale adjustment, f′ i,j,k is the eigenvalue of the (i, j, k)th position of the optimized training reference enhanced lithium battery detection feature map, and exp[·] represents the calculation of the natural exponential function value with the numerical value as the power.

[0021] In a second aspect, a lithium battery pole piece positioning system based on a dynamic filter is provided, which includes:

[0022] An image acquisition module, used to acquire a lithium battery image to be processed and a reference lithium battery image;

[0023] A first feature extraction module, used for passing the to-be-processed lithium battery image through a first convolutional neural network model as a feature extractor to obtain a lithium battery detection feature map;

[0024] A second feature extraction module, used to pass the reference lithium battery image through a second convolutional neural network model as a feature extractor to obtain a lithium battery reference feature map;

[0025] A spatial attention module, used for passing the lithium battery reference feature map through the spatial attention module to obtain a spatial attention weight matrix;

[0026] A calculation module, used for calculating the multiplication of each feature matrix along the channel dimension of the lithium battery detection feature map and the spatial attention weight matrix by position points to obtain a reference enhanced lithium battery detection feature map;

[0027] A dynamic filtering module, used for passing the reference enhanced lithium battery detection characteristic graph through the dynamic filtering module to obtain a filtered reference enhanced lithium battery detection characteristic graph; and

[0028] A decoding module is used to pass the filtered reference enhanced lithium battery detection feature map through a decoder to obtain a filtered lithium battery image.

[0029] In the above-mentioned dynamic filter-based lithium battery pole piece positioning system, the first feature extraction module is used to: use each layer of the second convolutional neural network model serving as a feature extractor to perform convolution processing, feature matrix-based mean pooling processing and nonlinear activation processing on the input data in the forward pass of the layer so that the last layer of the second convolutional neural network model serving as a feature extractor outputs the lithium battery detection feature map, wherein the input of the first layer of the second convolutional neural network model serving as a feature extractor is the lithium battery image to be processed.

[0030] In the above-mentioned dynamic filter-based lithium battery pole piece positioning system, the second feature extraction module is used to: use each layer of the second convolutional neural network model serving as a feature extractor to perform convolution processing, feature matrix-based mean pooling processing and nonlinear activation processing on the input data in the forward pass of the layer so that the last layer of the second convolutional neural network model serving as a feature extractor outputs the lithium battery reference feature map, wherein the input of the first layer of the second convolutional neural network model serving as a feature extractor is the reference lithium battery image.

[0031] Compared with the prior art, the lithium battery pole piece positioning method and system based on dynamic filter provided in the present application obtain the image of the lithium battery to be processed and the reference lithium battery image; and adopt artificial intelligence technology based on deep learning to spatially enhance and dynamically filter the image of the lithium battery to be processed, thereby realizing accurate positioning and identification of the positions of the positive and negative pole pieces. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0033] Figure 1 Schematic diagram of a lithium battery image type according to an embodiment of the present application.

[0034] Figure 2 The figure is a diagram of the overall network structure according to an embodiment of the present application.

[0035] Figure 3 Schematic diagram of the structure of a dynamic filtering module according to an embodiment of the present application.

[0036] Figure 4 It is a scene schematic diagram of the lithium battery pole piece positioning method based on the dynamic filter according to the embodiment of the present application.

[0037] Figure 5 This is a flow chart of a lithium battery pole piece positioning method based on a dynamic filter according to an embodiment of the present application.

[0038] Figure 6 Schematic diagram of the architecture of a lithium battery pole piece positioning method based on a dynamic filter according to an embodiment of the present application.

[0039] Figure 7 It is a flowchart of the sub-steps of step 140 in the lithium battery pole piece positioning method based on the dynamic filter according to an embodiment of the present application.

[0040] Figure 8 It is a flowchart of the sub-steps of step 180 in the lithium battery pole piece positioning method based on the dynamic filter according to an embodiment of the present application.

[0041] Fig. 9 It is a block diagram of a lithium battery pole piece positioning system based on a dynamic filter according to an embodiment of the present application. DETAILED DESCRIPTION

[0042] The following will describe the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0043] Unless otherwise specified, all technical and scientific terms used in the embodiments of the present application have the same meaning as those commonly understood by those skilled in the art of the present application. The terms used in this application are only for the purpose of describing specific embodiments and are not intended to limit the scope of this application.

[0044] In the embodiments of the present application, it should be noted that, unless otherwise specified and limited, the term "connection" should be understood in a broad sense. For example, it can be an electrical connection or a connection between two components. It can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to the specific circumstances.

[0045] It should be noted that the terms "first\second\third" involved in the embodiments of the present application are only used to distinguish similar objects, and do not represent a specific order for the objects. It is understandable that the specific order or sequence of "first\second\third" can be interchanged where permitted. It should be understood that the objects distinguished by "first\second\third" can be interchanged where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein.

[0046] Compared with fuel vehicles, when new energy vehicles first appeared, the most important issue for car owners was battery life. If the mobile phone is out of power, it can be turned on and used by charging it with a charging cable; if the new energy vehicle is out of power, it will end on the road. Therefore, at the beginning of the emergence of new energy vehicles, how to improve battery life became the primary problem faced by various car companies (at that time, due to the small base of new energy vehicles, battery safety issues had not yet fully emerged). From the 100-200km battery life represented by the small ant in the early stage, to the 300km-400km battery life represented by the BYD Qin / BAIC EU series recently, to the 500km-700km battery life represented by Weilai EC6 and Xiaopeng P7 in the past two years - in order to solve the "battery life anxiety" of new energy vehicle users, the R&D departments of major new energy vehicle companies are really working hard. Under this trend, the "high nickel" ternary lithium battery with the core orientation of improving energy density (i.e. battery life) has undoubtedly become the mainstream of the mainstream technology route. At the end of 2016, BYD, once the main advocate of lithium iron phosphate battery technology, announced that it would add 6GWh of ternary lithium battery production capacity in 2017. This was definitely explosive news for the lithium battery industry at the time, because BYD's move seemed to foreshadow the full retreat of the lithium iron phosphate battery route. In sharp contrast, CATL, the leader of "ternary lithium batteries", began to fully embrace "high-nickel" ternary lithium batteries in 2018 and became the first power battery manufacturer in the world to mass-produce high-nickel NCM 811 lithium batteries at that time. It was also from this year that ternary lithium batteries began to become the absolute overlord in the field of new energy vehicle power batteries with overwhelming advantages.

[0047] Since the X-ray of lithium batteries not only has inherent noise inside the equipment, but also has various forms of background interference, such as ear interference, diaphragm interference, tray distance too close, pole piece interference, etc. Therefore, it is very challenging to adapt to the detection model of various forms of interference at the same time. With the development of deep learning, the modeling methods and analogy tasks of industrial detection have become richer. Therefore, based on the image segmentation technology of deep learning, we model lithium battery detection as point segmentation, and introduce a dynamic filtering module to filter its interference and obtain discriminative positive and negative pole piece features.

[0048] The technical problems to be solved by this application are: 1. Providing a lithium battery detection model for general scenarios; 2. Overcoming background interference; 3. Accurate and stable performance.

[0049] In response to the above technical problems, the technical concept of the present application is: through deep learning-based image processing technology, the lithium battery image to be processed is spatially enhanced and dynamically filtered, so as to achieve accurate positioning and identification of the positive and negative electrode positions.

[0050] In this application, there are many interferences in the shooting of lithium batteries, such as Figure 1As shown in the figure, multiple types of interference pose a severe challenge to accurately locate the positive and negative electrodes. To this end, we propose an image segmentation algorithm based on dynamic filtering to improve the robustness and anti-interference ability of detection.

[0051] like Figure 2 As shown in the figure, an overall network structure is provided, which has two resnet-50 encoders, which extract the features of the current image and the features of the anchor as the standard filter respectively. We send the high-level three-layer current image and anchor features to the dynamic filter module for filtering, and send the output features to the decoder, and then the decoder and the encoder of the current image are continuously integrated to finally obtain the predicted polar mask.

[0052] like Figure 3 As shown in the figure, a dynamic filtering module is provided. First, the anchor feature is sent to the global average pooling (GAP) to obtain spatial attention, and then convolution and softmax are performed. After that, the filter parameters are generated by combining the convolution with the randomly initialized parameters, and then filtered with the features of the current image to obtain the filtered features.

[0053] The technical effects of the present application include: 1. A lithium battery pole piece positioning solution based on a dynamic filter is provided. 2. The above lithium battery pole piece positioning solution based on a dynamic filter performs spatial enhancement and dynamic filtering based on deep learning on the lithium battery image to be processed, thereby achieving accurate positioning and identification of the positive and negative pole pieces.

[0054] Specifically, in the technical solution of the present application, first, obtain the image of the lithium battery to be processed and the reference lithium battery image. In the lithium battery production process, in order to ensure the accurate positioning of the positive and negative electrodes, it is necessary to compare and detect each lithium battery image to be processed. Here, the reference lithium battery image refers to an image that can be used as a standard or benchmark to be compared with the lithium battery image to be processed, for example, a lithium battery image with a clear original image and no interference.

[0055] Taking into account the high complexity and variability of the lithium battery image to be processed, in the technical solution of the present application, the lithium battery image to be processed is passed through a first convolutional neural network model as a feature extractor for feature extraction and analysis, thereby obtaining a lithium battery detection feature map. Here, a convolutional neural network (CNN) is a deep learning algorithm that can automatically extract image features, and can gradually abstract and extract image features through multiple convolutional layers and pooling layers, such as edge, color, texture and other feature information in lithium battery images. Similarly, the reference lithium battery image is passed through a second convolutional neural network model as a feature extractor to extract high-dimensional implicit image features contained in the reference lithium battery image, thereby obtaining a lithium battery reference feature map.

[0056] In the positioning of the positive and negative electrodes of a lithium battery, a reference lithium battery image is used as a standard or benchmark image, which can be used to compare and detect the lithium battery image to be processed. In order to improve the accuracy and stability of positioning, in the technical solution of the present application, it is also expected that the lithium battery reference feature map will be subjected to information enhancement in the spatial dimension and elimination of interference information, that is, the lithium battery reference feature map is passed through a spatial attention module to adaptively adjust the weight coefficients of different areas in the feature map, so that the network ignores noise and useless information while paying attention to important positions, thereby obtaining a spatial attention weight matrix.

[0057] Then, the feature matrices along the channel dimension of the lithium battery detection feature map are calculated and multiplied by the spatial attention weight matrix by the position points to perform weighted processing on the feature matrices in the lithium battery detection feature map, that is, the spatial attention weight matrix is ​​multiplied as a weight matrix to obtain a reference enhanced lithium battery detection feature map. In this way, the key information in the lithium battery image is enhanced and highlighted.

[0058] Then, the reference enhanced lithium battery detection characteristic graph is passed through a dynamic filtering module to obtain a filtered reference enhanced lithium battery detection characteristic graph. The dynamic filtering module includes at least one convolution layer, which can capture features of different sizes based on convolution kernels of different scales, and dynamically adjust the filter parameters according to the current input conditions, thereby reducing noise, enhancing signals, etc. In this way, the robustness and stability of the reference enhanced lithium battery detection characteristic graph can be further improved, making it more adaptable to changes in different environments, thereby achieving more efficient and accurate positioning of positive and negative electrodes.

[0059] After obtaining the filtered reference enhanced lithium battery detection characteristic map, it is passed through a decoder to obtain a filtered lithium battery image. Here, the decoder can decode and regress the details and key information contained in the filtered reference enhanced lithium battery detection characteristic map based on the input filtered reference enhanced lithium battery detection characteristic map. In this way, an accurate and clear lithium battery image can be reconstructed, thereby realizing the precise positioning and identification of the positive and negative electrode positions.

[0060] Here, when calculating the position-based point multiplication of each feature matrix along the channel dimension of the lithium battery detection feature map and the spatial attention weight to obtain the reference enhanced lithium battery detection feature map, considering that the spatial attention weight is obtained by passing the lithium battery reference feature map through the spatial attention module, it essentially reflects the spatial image semantic distribution of the lithium battery reference feature map, which is different from the channel dimension feature distribution of the lithium battery detection feature map. Therefore, in order to improve the accuracy of the position-based point multiplication weighting, it is expected to improve the correlation between the spatial image semantic distribution of the lithium battery reference feature map and the channel dimension feature distribution of the lithium battery detection feature map.

[0061] Based on this, the applicant of this application considers improving the correlation between the spatial image semantic distribution of the lithium battery reference feature map and the channel dimension feature distribution of the lithium battery detection feature map by improving the overall feature learning correlation of the first convolutional neural network model, the second convolutional neural network model and the spatial attention module in the spatial dimension (i.e., the width dimension and height dimension of the feature matrix) and the channel dimension.

[0062] Specifically, during the training process, a Fourier-like scale domain probability correction is performed on the reference enhanced lithium battery detection characteristic map, for example, denoted as F, which is expressed as:

[0063]

[0064] Among them, f i,j,k ∈F is the eigenvalue of the (i, j, k)th position of the reference enhanced lithium battery detection feature map F, W, H and C are the height, width and number of channels of the reference enhanced lithium battery detection feature map F, respectively, and α, β and γ are hyperparameters for scale adjustment.

[0065] Here, the Fourier-like scale domain probability correction takes into account the homology between the high-dimensional feature distribution and the scale domain in which it is located, and can capture the potential distribution association in the homologous space based on the low-rank constraint of the scale space through the Fourier-like sparse low-rank transformation of the scale space. Therefore, during the training process of the first convolutional neural network model, the second convolutional neural network model and the spatial attention module, while extracting the features of the spatial dimension of the feature matrix and distributing the features on the channel dimension of the model, joint feature learning with the overall scale coherence of the feature map is achieved, so as to improve the expression accuracy of the reference enhanced lithium battery detection feature map by improving the learning correlation of the first convolutional neural network model, the second convolutional neural network model and the spatial attention module at the overall feature map expression scale, thereby improving the accuracy of the filtered lithium battery image obtained by the decoder of the filtered reference enhanced lithium battery detection feature map.

[0066] Figure 4 FIG. 1 is a schematic diagram of a scenario of a lithium battery pole piece positioning method based on a dynamic filter according to an embodiment of the present application. Figure 4 As shown, in this application scenario, first, obtain the lithium battery to be processed (for example, Figure 4 M) as shown in FIG. Figure 4 C1) and a reference lithium battery image (e.g., Figure 4 Then, the obtained lithium battery image to be processed and the reference lithium battery image are input to a server (for example, Figure 4 In S) shown in , the server is capable of processing the lithium battery image to be processed and the reference lithium battery image based on the lithium battery pole piece positioning algorithm of the dynamic filter to generate a filtered lithium battery image.

[0067] After introducing the basic principles of the present application, various non-limiting embodiments of the present application will be described in detail with reference to the accompanying drawings.

[0068] In one embodiment of the present application, Figure 5 FIG. 1 is a flow chart of a lithium battery pole piece positioning method based on a dynamic filter according to an embodiment of the present application. Figure 5As shown, according to the embodiment of the present application, a lithium battery pole piece positioning method 100 based on a dynamic filter includes: 110, obtaining a lithium battery image to be processed and a reference lithium battery image; 120, passing the lithium battery image to be processed through a first convolutional neural network model as a feature extractor to obtain a lithium battery detection feature map; 130, passing the reference lithium battery image through a second convolutional neural network model as a feature extractor to obtain a lithium battery reference feature map; 140, passing the lithium battery reference feature map through a spatial attention module to obtain a spatial attention weight matrix; 150, calculating each feature matrix of the lithium battery detection feature map along the channel dimension and multiplying the spatial attention weight matrix by position points to obtain a reference enhanced lithium battery detection feature map; 160, passing the reference enhanced lithium battery detection feature map through a dynamic filtering module to obtain a filtered reference enhanced lithium battery detection feature map; and, 170, passing the filtered reference enhanced lithium battery detection feature map through a decoder to obtain a filtered lithium battery image.

[0069] Figure 6 FIG. 1 is a schematic diagram of the structure of a lithium battery pole piece positioning method based on a dynamic filter according to an embodiment of the present application. Figure 6 As shown, in the network architecture, first, a lithium battery image to be processed and a reference lithium battery image are obtained; then, the lithium battery image to be processed is passed through a first convolutional neural network model as a feature extractor to obtain a lithium battery detection feature map; then, the reference lithium battery image is passed through a second convolutional neural network model as a feature extractor to obtain a lithium battery reference feature map; then, the lithium battery reference feature map is passed through a spatial attention module to obtain a spatial attention weight matrix; then, each feature matrix along the channel dimension of the lithium battery detection feature map is multiplied by the spatial attention weight matrix by a position point to obtain a reference enhanced lithium battery detection feature map; then, the reference enhanced lithium battery detection feature map is passed through a dynamic filtering module to obtain a filtered reference enhanced lithium battery detection feature map; and, finally, the filtered reference enhanced lithium battery detection feature map is passed through a decoder to obtain a filtered lithium battery image.

[0070] Specifically, in step 110, the lithium battery image to be processed and the reference lithium battery image are obtained. In view of the above technical problems, the technical concept of the present application is: through the image processing technology based on deep learning, the lithium battery image to be processed is spatially enhanced and dynamically filtered, so as to achieve accurate positioning and identification of the positive and negative electrode positions.

[0071] Specifically, in the technical solution of the present application, first, obtain the image of the lithium battery to be processed and the reference lithium battery image. In the lithium battery production process, in order to ensure the accurate positioning of the positive and negative electrodes, it is necessary to compare and detect each lithium battery image to be processed. Here, the reference lithium battery image refers to an image that can be used as a standard or benchmark to be compared with the lithium battery image to be processed, for example, a lithium battery image with a clear original image and no interference.

[0072] Specifically, in step 120 and step 130, the lithium battery image to be processed is passed through the first convolutional neural network model as a feature extractor to obtain a lithium battery detection feature map; and the reference lithium battery image is passed through the second convolutional neural network model as a feature extractor to obtain a lithium battery reference feature map. Considering that the lithium battery image to be processed has high complexity and variability, in the technical solution of the present application, the lithium battery image to be processed is passed through the first convolutional neural network model as a feature extractor to perform feature extraction and analysis, thereby obtaining a lithium battery detection feature map.

[0073] Here, the convolutional neural network (CNN) is a deep learning algorithm that can automatically extract image features, and can gradually abstract and extract image features through multiple convolutional layers and pooling layers, such as edge, color, texture and other feature information in lithium battery images. Similarly, the reference lithium battery image is passed through a second convolutional neural network model as a feature extractor to extract high-dimensional implicit image features contained in the reference lithium battery image, thereby obtaining a lithium battery reference feature map.

[0074] Wherein, the lithium battery image to be processed is passed through the first convolutional neural network model as a feature extractor to obtain a lithium battery detection feature map, including: using each layer of the second convolutional neural network model as a feature extractor to perform convolution processing, feature matrix-based mean pooling processing and nonlinear activation processing on the input data in the forward transfer of the layer so that the last layer of the second convolutional neural network model as a feature extractor outputs the lithium battery detection feature map, wherein the input of the first layer of the second convolutional neural network model as a feature extractor is the lithium battery image to be processed.

[0075] The reference lithium battery image is passed through a second convolutional neural network model as a feature extractor to obtain a lithium battery reference feature map, including: using each layer of the second convolutional neural network model as a feature extractor to perform convolution processing, feature matrix-based mean pooling processing and nonlinear activation processing on the input data in the forward transfer of the layer so that the last layer of the second convolutional neural network model as a feature extractor outputs the lithium battery reference feature map, wherein the input of the first layer of the second convolutional neural network model as a feature extractor is the reference lithium battery image.

[0076] Convolutional Neural Network (CNN) is an artificial neural network that is widely used in image recognition and other fields. Convolutional neural networks can include input layers, hidden layers, and output layers, where the hidden layers can include convolutional layers, pooling layers, activation layers, and fully connected layers. The previous layer performs corresponding operations based on the input data and outputs the operation results to the next layer. The initial input data is processed through multiple layers of operations to obtain a final result.

[0077] The convolutional neural network model uses the convolution kernel as a feature filtering factor and has excellent performance in extracting local features of images. Compared with traditional image feature extraction algorithms based on statistics or feature engineering, the convolutional neural network model has stronger feature extraction generalization and fitting capabilities.

[0078] Specifically, in step 140, the lithium battery reference feature map is passed through a spatial attention module to obtain a spatial attention weight matrix. In the positioning of the positive and negative electrodes of a lithium battery, the reference lithium battery image is used as a standard or benchmark image, which can be used to compare and detect the lithium battery image to be processed. In order to improve the accuracy and stability of positioning, in the technical solution of the present application, it is also expected that the lithium battery reference feature map is enhanced in the spatial dimension and the interference information is eliminated, that is, the lithium battery reference feature map is passed through a spatial attention module to adaptively adjust the weight coefficients of different areas in the feature map, so that the network ignores noise and useless information while paying attention to important positions, thereby obtaining a spatial attention weight matrix.

[0079] Figure 7 Flow chart of the sub-steps of step 140 in the lithium battery pole piece positioning method based on a dynamic filter according to an embodiment of the present application, as shown in FIG. Figure 7As shown, the lithium battery reference feature map is passed through a spatial attention module to obtain a spatial attention weight matrix, including: 141, calculating the global mean of the lithium battery reference feature map along the channel dimension to obtain a lithium battery reference feature matrix; 142, passing the lithium battery reference feature matrix through a convolutional layer to obtain a spatial attention feature matrix; and, 143, inputting the spatial attention feature matrix into a Softmax activation function to obtain the spatial attention weight feature matrix.

[0080] The attention mechanism is a data processing method in machine learning, which is widely used in various types of machine learning tasks such as natural language processing, image recognition and speech recognition. On the one hand, the attention mechanism hopes that the network can automatically learn the areas that need attention in the picture or text sequence; on the other hand, the attention mechanism generates a mask through the operation of the neural network, and the weight of the value on the mask. Generally speaking, the spatial attention mechanism calculates the average of different channels of the same pixel, and then obtains the spatial features through some convolution and upsampling operations. The pixels of each layer of the spatial features are given different weights.

[0081] Specifically, in step 150, each feature matrix of the lithium battery detection feature map along the channel dimension is multiplied by the position point of the spatial attention weight matrix to obtain a reference enhanced lithium battery detection feature map. Then, each feature matrix of the lithium battery detection feature map along the channel dimension is multiplied by the position point of the spatial attention weight matrix to perform weighted processing on each feature matrix in the lithium battery detection feature map, that is, the spatial attention weight matrix is ​​multiplied as a weight matrix to obtain a reference enhanced lithium battery detection feature map. In this way, the key information in the lithium battery image is enhanced and highlighted.

[0082] Specifically, in step 160, the reference enhanced lithium battery detection characteristic graph is passed through a dynamic filtering module to obtain a filtered reference enhanced lithium battery detection characteristic graph. Subsequently, the reference enhanced lithium battery detection characteristic graph is passed through a dynamic filtering module to obtain a filtered reference enhanced lithium battery detection characteristic graph. Among them, the dynamic filtering module includes at least one convolution layer, and the convolution layer can capture features of different sizes based on convolution kernels of different scales, and dynamically adjust the filter parameters according to the current input situation, thereby reducing noise, enhancing signals, etc. In this way, the robustness and stability of the reference enhanced lithium battery detection characteristic graph can be further improved, making it more adaptable to changes in different environments, thereby achieving more efficient and accurate positioning of the positive and negative electrodes.

[0083] Specifically, in step 170, the filtered reference enhanced lithium battery detection feature map is passed through a decoder to obtain a filtered lithium battery image. After obtaining the filtered reference enhanced lithium battery detection feature map, it is passed through a decoder to obtain a filtered lithium battery image. Here, the decoder can decode and regress the details and key information contained in the filtered reference enhanced lithium battery detection feature map based on the input filtered reference enhanced lithium battery detection feature map. In this way, an accurate and clear lithium battery image can be reconstructed, thereby achieving precise positioning and identification of the positive and negative electrode positions.

[0084] In a specific example of the present application, the decoder is used to decode and regress the filtered reference enhanced lithium battery detection characteristic diagram using the following formula to obtain the decoded value; wherein the formula is: Among them, F d represents the reference enhanced lithium battery detection feature map after filtering, Y represents the lithium battery image after filtering, W represents the weight matrix, B represents the bias vector, Represents matrix multiplication.

[0085] The lithium battery pole piece positioning method based on the dynamic filter also includes a training step: training the first convolutional neural network model as a feature extractor, the second convolutional neural network model as a feature extractor, the spatial attention module, the dynamic filtering module and the decoder; Figure 8 Flow chart of the sub-steps of step 180 in the lithium battery pole piece positioning method based on a dynamic filter according to an embodiment of the present application, such as Figure 8As shown, the training step 180 includes: 1801, obtaining training data, the training data including a to-be-processed training lithium battery image and a training reference lithium battery image, as well as a real filtered lithium battery image; 1802, passing the to-be-processed training lithium battery image through the first convolutional neural network model as a feature extractor to obtain a training lithium battery detection feature map; 1803, passing the training reference lithium battery image through the second convolutional neural network model as a feature extractor to obtain a training lithium battery reference feature map; 1804, passing the training lithium battery reference feature map through the spatial attention module to obtain a training spatial attention weight matrix; 1805, calculating the position point multiplication of each feature matrix of the training lithium battery detection feature map along the channel dimension and the training spatial attention weight matrix to obtain a training reference enhanced lithium battery detection feature map; 1806, for the training The training reference enhanced lithium battery detection feature map is subjected to Fourier scale domain probability correction to obtain an optimized training reference enhanced lithium battery detection feature map; 1807, the optimized training reference enhanced lithium battery detection feature map is passed through the dynamic filtering module to obtain a training filtered reference enhanced lithium battery detection feature map; 1808, the training filtered reference enhanced lithium battery detection feature map is passed through the decoder to obtain a training filtered lithium battery image; 1809, the mean square error between the training filtered lithium battery image and the true filtered lithium battery image is calculated as a decoding loss function value; and, 1810, based on the decoding loss function value and through the direction propagation of gradient descent, the first convolutional neural network model as a feature extractor, the second convolutional neural network model as a feature extractor, the spatial attention module, the dynamic filtering module and the decoder are trained.

[0086] Here, when calculating the position-based point multiplication of each feature matrix along the channel dimension of the lithium battery detection feature map and the spatial attention weight to obtain the reference enhanced lithium battery detection feature map, considering that the spatial attention weight is obtained by passing the lithium battery reference feature map through the spatial attention module, it essentially reflects the spatial image semantic distribution of the lithium battery reference feature map, which is different from the channel dimension feature distribution of the lithium battery detection feature map. Therefore, in order to improve the accuracy of the position-based point multiplication weighting, it is expected to improve the correlation between the spatial image semantic distribution of the lithium battery reference feature map and the channel dimension feature distribution of the lithium battery detection feature map.

[0087] Based on this, the applicant of this application considers improving the correlation between the spatial image semantic distribution of the lithium battery reference feature map and the channel dimension feature distribution of the lithium battery detection feature map by improving the overall feature learning correlation of the first convolutional neural network model, the second convolutional neural network model and the spatial attention module in the spatial dimension (i.e., the width dimension and height dimension of the feature matrix) and the channel dimension.

[0088] Specifically, during the training process, the reference enhanced lithium battery detection feature map, for example, denoted as F, is subjected to a Fourier scale domain probability correction, which is expressed as: the training reference enhanced lithium battery detection feature map is subjected to a Fourier scale domain probability correction using the following optimization formula to obtain an optimized training reference enhanced lithium battery detection feature map; wherein the optimization formula is:

[0089]

[0090] Wherein, F represents the reference enhanced lithium battery detection characteristic diagram, f i,j,k ∈F is the eigenvalue of the (i, j, k)th position of the reference enhanced lithium battery detection feature map, W, H and C are the height, width and number of channels of the reference enhanced lithium battery detection feature map, respectively, and α, β and γ are hyperparameters for scale adjustment, f′ i,j,k is the eigenvalue of the (i, j, k)th position of the optimized training reference enhanced lithium battery detection feature map, and exp[·] represents the calculation of the natural exponential function value with the numerical value as the power.

[0091] Here, the Fourier-like scale domain probability correction takes into account the homology between the high-dimensional feature distribution and the scale domain in which it is located, and can capture the potential distribution association in the homologous space based on the low-rank constraint of the scale space through the Fourier-like sparse low-rank transformation of the scale space. Therefore, during the training process of the first convolutional neural network model, the second convolutional neural network model and the spatial attention module, while extracting the features of the spatial dimension of the feature matrix and distributing the features on the channel dimension of the model, joint feature learning with the overall scale coherence of the feature map is achieved, so as to improve the expression accuracy of the reference enhanced lithium battery detection feature map by improving the learning correlation of the first convolutional neural network model, the second convolutional neural network model and the spatial attention module at the overall feature map expression scale, thereby improving the accuracy of the filtered lithium battery image obtained by the decoder of the filtered reference enhanced lithium battery detection feature map.

[0092] In summary, a lithium battery pole piece positioning method 100 based on a dynamic filter according to an embodiment of the present application is explained, which obtains a lithium battery image to be processed and a reference lithium battery image; and uses artificial intelligence technology based on deep learning to spatially enhance and dynamically filter the lithium battery image to be processed, thereby achieving accurate positioning and identification of the positions of the positive and negative pole pieces.

[0093] In one embodiment of the present application, Fig. 9 FIG. 1 is a block diagram of a lithium battery pole piece positioning system based on a dynamic filter according to an embodiment of the present application. Fig. 9 As shown, according to the embodiment of the present application, the lithium battery pole piece positioning system 200 based on the dynamic filter includes: an image acquisition module 210, which is used to acquire a lithium battery image to be processed and a reference lithium battery image; a first feature extraction module 220, which is used to pass the lithium battery image to be processed through a first convolutional neural network model as a feature extractor to obtain a lithium battery detection feature map; a second feature extraction module 230, which is used to pass the reference lithium battery image through a second convolutional neural network model as a feature extractor to obtain a lithium battery reference feature map; a spatial attention module 240, which is used to pass the lithium battery reference feature map through the spatial attention module to obtain a spatial attention weight matrix; a calculation module 250, which is used to calculate the multiplication of each feature matrix along the channel dimension of the lithium battery detection feature map and the spatial attention weight matrix by position point to obtain a reference enhanced lithium battery detection feature map; a dynamic filtering module 260, which is used to pass the reference enhanced lithium battery detection feature map through the dynamic filtering module to obtain a filtered reference enhanced lithium battery detection feature map; and a decoding module 270, which is used to pass the filtered reference enhanced lithium battery detection feature map through a decoder to obtain a filtered lithium battery image.

[0094] In a specific example, in the above-mentioned dynamic filter-based lithium battery pole piece positioning system, the first feature extraction module is used to: use each layer of the second convolutional neural network model serving as a feature extractor to perform convolution processing, feature matrix-based mean pooling processing and nonlinear activation processing on the input data in the forward pass of the layer so that the last layer of the second convolutional neural network model serving as a feature extractor outputs the lithium battery detection feature map, wherein the input of the first layer of the second convolutional neural network model serving as a feature extractor is the lithium battery image to be processed.

[0095] In a specific example, in the above-mentioned dynamic filter-based lithium battery pole piece positioning system, the second feature extraction module is used to: use each layer of the second convolutional neural network model serving as a feature extractor to perform convolution processing, feature matrix-based mean pooling processing and nonlinear activation processing on the input data in the forward pass of the layer so that the last layer of the second convolutional neural network model serving as a feature extractor outputs the lithium battery reference feature map, wherein the input of the first layer of the second convolutional neural network model serving as a feature extractor is the reference lithium battery image.

[0096] In a specific example, in the above-mentioned dynamic filter-based lithium battery pole piece positioning system, the spatial attention module includes: a global mean unit, used to calculate the global mean of the lithium battery reference feature map along the channel dimension to obtain a lithium battery reference feature matrix; a convolution unit, used to pass the lithium battery reference feature matrix through a convolution layer to obtain a spatial attention feature matrix; and an activation unit, used to input the spatial attention feature matrix into a Softmax activation function to obtain the spatial attention weight feature matrix.

[0097] In a specific example, in the above-mentioned lithium battery pole piece positioning system based on dynamic filter, the dynamic filtering module includes at least one convolution layer.

[0098] In a specific example, in the above-mentioned lithium battery pole piece positioning system based on dynamic filter, a training step is also included: a training module for training the first convolutional neural network model as a feature extractor, the second convolutional neural network model as a feature extractor, the spatial attention module, the dynamic filtering module and the decoder; wherein the training module includes: a training image acquisition unit for acquiring training data, the training data including a training lithium battery image to be processed and a training reference lithium battery image, as well as a real filtered lithium battery image; a training first feature extraction unit for passing the training lithium battery image to be processed through the first convolutional neural network model as a feature extractor to obtain a training lithium battery detection feature map; a training second feature extraction unit for passing the training reference lithium battery image through the second convolutional neural network model as a feature extractor to obtain a training lithium battery reference feature map; a training spatial attention unit for passing the training lithium battery reference feature map through the spatial attention module to obtain a training spatial attention weight matrix; a training calculation module unit for calculating the training lithium battery detection feature map. Each feature matrix along the channel dimension of the graph is multiplied by the training spatial attention weight matrix at each position point to obtain a training reference enhanced lithium battery detection feature map; a training optimization unit is used to perform a Fourier-like scale domain probability correction on the training reference enhanced lithium battery detection feature map to obtain an optimized training reference enhanced lithium battery detection feature map; a training dynamic filtering unit is used to pass the optimized training reference enhanced lithium battery detection feature map through the dynamic filtering module to obtain a training filtered reference enhanced lithium battery detection feature map; a training decoding unit is used to pass the training filtered reference enhanced lithium battery detection feature map through the decoder to obtain a training filtered lithium battery image; a decoding loss function value calculation unit is used to calculate the mean square error between the training filtered lithium battery image and the real filtered lithium battery image as a decoding loss function value; and a training unit is used to train the first convolutional neural network model as a feature extractor, the second convolutional neural network model as a feature extractor, the spatial attention module, the dynamic filtering module and the decoder based on the decoding loss function value and through the direction propagation of gradient descent.

[0099] In a specific example, in the above-mentioned lithium battery pole piece positioning system based on the dynamic filter, the training optimization unit is used to: perform a Fourier scale domain probability correction on the training reference enhanced lithium battery detection characteristic diagram using the following optimization formula to obtain an optimized training reference enhanced lithium battery detection characteristic diagram; wherein the optimization formula is:

[0100]

[0101] Wherein, F represents the reference enhanced lithium battery detection characteristic diagram, f i,j,k ∈F is the eigenvalue of the (i, j, k)th position of the reference enhanced lithium battery detection feature map, W, H and C are the height, width and number of channels of the reference enhanced lithium battery detection feature map, respectively, and α, β and γ are hyperparameters for scale adjustment, f′ i,j,k is the eigenvalue of the (i, j, k)th position of the optimized training reference enhanced lithium battery detection feature map, and exp[·] represents the calculation of the natural exponential function value with the numerical value as the power.

[0102] Here, those skilled in the art can understand that the specific functions and operations of each unit and module in the above-mentioned lithium battery pole piece positioning system based on dynamic filter have been referred to above. Figures 1 to 8 The description of the lithium battery pole piece positioning method based on dynamic filter has been introduced in detail, and therefore, its repeated description will be omitted.

[0103] As described above, the lithium battery pole piece positioning system 200 based on the dynamic filter according to the embodiment of the present application can be implemented in various terminal devices, such as a server for positioning the lithium battery pole piece based on the dynamic filter. In one example, the lithium battery pole piece positioning system 200 based on the dynamic filter according to the embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the lithium battery pole piece positioning system 200 based on the dynamic filter can be a software module in the operating system of the terminal device, or it can be an application developed for the terminal device; of course, the lithium battery pole piece positioning system 200 based on the dynamic filter can also be one of the many hardware modules of the terminal device.

[0104] Alternatively, in another example, the dynamic filter-based lithium battery pole piece positioning system 200 and the terminal device may also be separate devices, and the dynamic filter-based lithium battery pole piece positioning system 200 may be connected to the terminal device via a wired and / or wireless network, and transmit interactive information in accordance with an agreed data format.

[0105] The present application also provides a computer program product, which includes instructions. When the instructions are executed, the device performs operations corresponding to the above method.

[0106] In one embodiment of the present application, a computer-readable storage medium is further provided, wherein the computer-readable storage medium stores a computer program for executing the above-mentioned method.

[0107] It should be understood that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware can be adopted. Moreover, 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 can be adopted.

[0108] The method, system, and computer program product of the embodiments of the present application are described by flowcharts and / or block diagrams. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 one or more processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0109] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 one or more processes and / or boxes Figure 1 A function specified in one or more boxes.

[0110] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 one or more processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0111] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details disclosed above are only for the purpose of illustration and ease of understanding, not for limitation, and the above details do not limit the present application to being implemented by adopting the above specific details.

[0112] The block diagrams of the devices, apparatuses, equipment, and systems involved in this application are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagram. As will be appreciated by those skilled in the art, these devices, apparatuses, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including", "comprising", "having", etc. are open words, referring to "including but not limited to", and can be used interchangeably with them. The words "or" and "and" used here refer to the words "and / or" and can be used interchangeably with them, unless the context clearly indicates otherwise. The words "such as" used here refer to the phrase "such as but not limited to", and can be used interchangeably with them.

[0113] It should also be noted that in the apparatus, device and method of the present application, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present application.

[0114] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

[0115] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.

[0116] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. A lithium battery pole piece positioning method based on a dynamic filter, characterized in that: include: Obtaining a lithium battery image to be processed and a reference lithium battery image, wherein, in order to ensure accurate positioning of the positive and negative electrode sheets, each lithium battery image to be processed needs to be compared and detected, wherein the reference lithium battery image refers to an image that can be used as a standard or benchmark for comparison with the lithium battery image to be processed; Passing the to-be-processed lithium battery image through a first convolutional neural network model as a feature extractor to obtain a lithium battery detection feature map; Passing the reference lithium battery image through a second convolutional neural network model as a feature extractor to obtain a lithium battery reference feature map; Pass the lithium battery reference feature map through a spatial attention module to obtain a spatial attention weight matrix; Calculate the point-by-point multiplication of each feature matrix along the channel dimension of the lithium battery detection feature map and the spatial attention weight matrix to obtain a reference enhanced lithium battery detection feature map; Passing the reference enhanced lithium battery detection characteristic graph through a dynamic filtering module to obtain a filtered reference enhanced lithium battery detection characteristic graph, wherein the dynamic filtering module includes at least one convolution layer, and the convolution layer can capture features of different sizes based on convolution kernels of different scales, and dynamically adjust filter parameters according to current input conditions, thereby reducing noise and enhancing signals; and The filtered reference enhanced lithium battery detection characteristic image is passed through a decoder to obtain a filtered lithium battery image.

2. The lithium battery pole piece positioning method based on dynamic filter according to claim 1 is characterized in that: The lithium battery image to be processed is passed through the first convolutional neural network model as a feature extractor to obtain a lithium battery detection feature map, including: using each layer of the first convolutional neural network model as a feature extractor to perform convolution processing, feature matrix-based mean pooling processing and nonlinear activation processing on the input data in the forward transfer of the layer so that the last layer of the first convolutional neural network model as a feature extractor outputs the lithium battery detection feature map, wherein the input of the first layer of the first convolutional neural network model as a feature extractor is the lithium battery image to be processed.

3. The lithium battery pole piece positioning method based on dynamic filter according to claim 2 is characterized in that: The reference lithium battery image is passed through a second convolutional neural network model as a feature extractor to obtain a lithium battery reference feature map, including: using each layer of the second convolutional neural network model as a feature extractor to perform convolution processing, feature matrix-based mean pooling processing and nonlinear activation processing on the input data in the forward transfer of the layer so that the last layer of the second convolutional neural network model as a feature extractor outputs the lithium battery reference feature map, wherein the input of the first layer of the second convolutional neural network model as a feature extractor is the reference lithium battery image.

4. The lithium battery pole piece positioning method based on dynamic filter according to claim 3 is characterized in that: The lithium battery reference feature map is passed through a spatial attention module to obtain a spatial attention weight matrix, including: Calculating the global mean of the lithium battery reference characteristic graph along the channel dimension to obtain a lithium battery reference characteristic matrix; Passing the lithium battery reference feature matrix through a convolutional layer to obtain a spatial attention feature matrix; and The spatial attention feature matrix is ​​input into the Softmax activation function to obtain the spatial attention weight feature matrix.

5. The lithium battery pole piece positioning method based on dynamic filter according to claim 4 is characterized in that: The method further includes a training step of: training the first convolutional neural network model as a feature extractor, the second convolutional neural network model as a feature extractor, the spatial attention module, the dynamic filtering module and the decoder; Wherein, the training step includes: Acquire training data, wherein the training data includes a to-be-processed training lithium battery image and a training reference lithium battery image, as well as a real filtered lithium battery image; Passing the to-be-processed training lithium battery image through the first convolutional neural network model as a feature extractor to obtain a training lithium battery detection feature map; Passing the training reference lithium battery image through the second convolutional neural network model as a feature extractor to obtain a training lithium battery reference feature map; Passing the training lithium battery reference feature map through the spatial attention module to obtain a training spatial attention weight matrix; Calculate the point-by-point multiplication of each feature matrix along the channel dimension of the training lithium battery detection feature map and the training spatial attention weight matrix to obtain a training reference enhanced lithium battery detection feature map; Performing Fourier-like scale domain probability correction on the training reference enhanced lithium battery detection feature map to obtain an optimized training reference enhanced lithium battery detection feature map; Passing the optimized training reference enhanced lithium battery detection characteristic graph through the dynamic filtering module to obtain a training filtered reference enhanced lithium battery detection characteristic graph; Passing the trained filtered reference enhanced lithium battery detection feature map through the decoder to obtain a trained filtered lithium battery image; Calculating a mean square error between the training filtered lithium battery image and the real filtered lithium battery image as a decoding loss function value; and The first convolutional neural network model as a feature extractor, the second convolutional neural network model as a feature extractor, the spatial attention module, the dynamic filtering module and the decoder are trained based on the decoding loss function value and propagated in the direction of gradient descent.

6. The lithium battery pole piece positioning method based on dynamic filter according to claim 5 is characterized in that: The method further comprises: performing a Fourier-like scale domain probability correction on the training reference enhanced lithium battery detection characteristic graph to obtain an optimized training reference enhanced lithium battery detection characteristic graph, comprising: performing a Fourier-like scale domain probability correction on the training reference enhanced lithium battery detection characteristic graph to obtain an optimized training reference enhanced lithium battery detection characteristic graph according to the following optimization formula; Wherein, the optimization formula is: Wherein, F represents the reference enhanced lithium battery detection characteristic diagram, f i，j，k ∈F is the eigenvalue of the (i, j, k)th position of the reference enhanced lithium battery detection feature map, W, H and C are the height, width and number of channels of the reference enhanced lithium battery detection feature map, respectively, and α, β and γ are hyperparameters for scale adjustment, f′ i，j，k is the eigenvalue of the (i, j, k)th position of the optimized training reference enhanced lithium battery detection feature map, and exp[·] represents the calculation of the natural exponential function value with the numerical value as the power.

7. A lithium battery pole piece positioning system based on a dynamic filter, characterized in that: include: An image acquisition module is used to acquire a to-be-processed lithium battery image and a reference lithium battery image, wherein, in order to ensure accurate positioning of the positive and negative electrode sheets, each to-be-processed lithium battery image needs to be compared and detected, wherein the reference lithium battery image refers to an image that can be used as a standard or benchmark for comparison with the to-be-processed lithium battery image; A first feature extraction module, used for passing the to-be-processed lithium battery image through a first convolutional neural network model as a feature extractor to obtain a lithium battery detection feature map; A second feature extraction module, used to pass the reference lithium battery image through a second convolutional neural network model as a feature extractor to obtain a lithium battery reference feature map; A spatial attention module, used for passing the lithium battery reference feature map through the spatial attention module to obtain a spatial attention weight matrix; A calculation module, used for calculating the multiplication of each feature matrix along the channel dimension of the lithium battery detection feature map and the spatial attention weight matrix by position points to obtain a reference enhanced lithium battery detection feature map; A dynamic filtering module, used for passing the reference enhanced lithium battery detection characteristic graph through the dynamic filtering module to obtain a filtered reference enhanced lithium battery detection characteristic graph, wherein the dynamic filtering module includes at least one convolution layer, and the convolution layer can capture features of different sizes based on convolution kernels of different scales, and dynamically adjust the filter parameters according to the current input situation, so as to reduce noise and enhance the signal; and A decoding module is used to pass the filtered reference enhanced lithium battery detection feature map through a decoder to obtain a filtered lithium battery image.

8. The lithium battery pole piece positioning system based on dynamic filter according to claim 7 is characterized in that: The first feature extraction module is used to: use each layer of the first convolutional neural network model as a feature extractor to perform convolution processing, feature matrix-based mean pooling processing and nonlinear activation processing on the input data in the forward pass of the layer so that the last layer of the first convolutional neural network model as a feature extractor outputs the lithium battery detection feature map, wherein the input of the first layer of the first convolutional neural network model as a feature extractor is the lithium battery image to be processed.

9. The lithium battery pole piece positioning system based on dynamic filter according to claim 8, characterized in that: The second feature extraction module is used to: use each layer of the second convolutional neural network model as a feature extractor to perform convolution processing, feature matrix-based mean pooling processing and nonlinear activation processing on the input data in the forward pass of the layer so that the last layer of the second convolutional neural network model as a feature extractor outputs the lithium battery reference feature map, wherein the input of the first layer of the second convolutional neural network model as a feature extractor is the reference lithium battery image.

Images