A pole piece detection method and device, electronic equipment and storage medium

Abstract

This invention provides an electrode detection method, apparatus, electronic device, and storage medium, relating to the field of battery testing technology. The electrode detection method of this invention includes: setting a detection box in the image of the electrode to be inspected; extracting image features within the detection box, wherein the detection box includes an edge damage detection box for detecting corner breakage of the electrode and an electrode tab folding detection box for detecting tab folding defects; inputting the image features into a preset artificial intelligence model, and outputting the electrode detection result, wherein the preset artificial intelligence model is based on a multi-layer convolutional layer and fully connected layer structure, and optimizes the model parameters through deep learning on different electrode samples. This invention achieves automated identification and classification of electrode appearance defects, can automatically extract image features at different levels, adapts to irregular damage and complex lighting conditions, thereby effectively improving the generalization ability and accuracy of detection.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of battery detection, in particular to a pole piece detection method and device, electronic equipment and storage medium. BACKGROUND

[0002] In the battery manufacturing process, as a key component, the surface defects of the pole piece, such as uneven coating, scratches, foreign matter, damage, and tab folding, may all cause safety hazards such as battery performance degradation, short circuit, or thermal runaway. Therefore, a visual detection system is usually configured in the pole piece cutting and stacking or winding process to automatically detect and judge the appearance defects of the pole piece.

[0003] In related technologies, the detection logic of the pole piece defects is mainly based on traditional image processing methods, which analyzes the image gray scale distribution, calculates the standard deviation or local thickness deviation, and combines the set threshold to judge the defects. Although preliminary automatic detection is achieved, there are still problems such as misidentification and missed killing, which seriously affect the process yield and product quality and safety. SUMMARY

[0004] The problem solved by the present application is how to improve the detection accuracy of the pole piece.

[0005] To solve the above problems, the present application provides a pole piece detection method, device, electronic equipment and storage medium.

[0006] In a first aspect, the present application provides a pole piece detection method, comprising: setting a detection frame in a to-be-detected image of a pole piece, and extracting image features in the detection frame, wherein the detection frame includes an edge damage detection frame for detecting edge corner damage of the pole piece and a tab folding detection frame for detecting tab folding defects. inputting the image features into a preset artificial intelligence model to output a pole piece detection result, wherein the preset artificial intelligence model is based on a multi-layer convolution layer and a full connection layer structure, and the model parameters are optimized through deep learning of different pole piece samples.

[0007] Optionally, the step of setting a detection frame in a to-be-detected image of a pole piece and extracting image features in the detection frame includes: setting the edge damage detection frame in the corner area of the to-be-detected image; extracting edge pixel points of the to-be-detected image from the edge damage detection frame as the image features.

[0008] Optionally, the step of extracting edge pixel points of the to-be-detected image from the edge damage detection frame as the image features includes: determining the accurate position of the pole piece to establish a local coordinate axis through horizontal edge grabbing and vertical edge grabbing; extract the edge pixel point based on the local coordinate axis.

[0009] Optionally, the inputting the image feature into the preset artificial intelligence model to output a pole piece detection result comprises: inputting the edge pixel point into the preset artificial intelligence model, extracting hierarchical features of the edge corner region by using a convolution layer, and realizing damage classification by using a full connection layer to output a pole piece edge corner damage state, wherein the pole piece edge corner damage state comprises a damage category and a position coordinate.

[0010] Optionally, after the extracting the edge pixel point of the to-be-detected image from the edge damage detection frame, the pole piece detection method further comprises: fitting the coordinates of the edge pixel point to determine a fitting curve of the damaged edge, comparing the fitting curve with a corresponding local coordinate axis, and determining whether the pole piece edge corner is a real damage according to an offset or an angle deviation between the fitting curve and the local coordinate axis.

[0011] Optionally, the setting a detection frame in the to-be-detected image of a pole piece and extracting image features in the detection frame comprises: setting the pole lug folding detection frame in the pole lug region of the to-be-detected image, wherein the pole lug folding detection frame can be regionally adjusted; extracting image features of the to-be-detected image from the pole lug folding detection frame.

[0012] Optionally, the extracting image features of the to-be-detected image from the pole lug folding detection frame comprises: performing gray scale processing on the image in the pole lug folding detection frame to determine a suspected defect position; performing gray scale erosion on the image in the region where the suspected defect position is located, and extracting the image features after morphological conversion of the image processed by the gray scale erosion.

[0013] Optionally, the inputting the image feature into the preset artificial intelligence model to output a pole piece detection result comprises: inputting the image features processed by the gray scale erosion and the morphological conversion into the preset artificial intelligence model, extracting hierarchical features of the pole lug region by using a convolution layer, and realizing defect classification by using a full connection layer to output a pole lug defect state, wherein the pole lug defect state comprises a defect category and a position coordinate.

[0014] Optionally, the preset artificial intelligence model adopts a multi-task deep learning model, which includes a shared feature extraction network and multiple sub-models. The shared feature extraction network is used to perform unified hierarchical feature extraction on the input image features, and the multiple sub-models are respectively used to perform different types of defect detection tasks based on the output of the shared feature extraction network.

[0015] Optionally, the sub-model includes a first sub-model for detecting electrode corner damage. The first sub-model analyzes the edge pixel features of the electrode corner region based on the feature information output by the shared feature extraction network, classifies the edge pixel features for damage, and outputs the electrode corner damage status.

[0016] Optionally, the sub-model includes a second sub-model for detecting tab folding defects. The second sub-model analyzes the image features of the tab region based on the feature information output by the shared feature extraction network, classifies the image features of the tab region as defects, and outputs the tab defect status.

[0017] Optionally, the multi-task deep learning model further includes a task collaborative judgment module, which is used to perform correlation analysis on the electrode edge damage state or electrode tab defect state output by the sub-model. Based on the correlation of different defect types in spatial location, morphological characteristics or occurrence probability, the detection results of the sub-model are comprehensively corrected or the confidence level is adjusted to output the final electrode detection result.

[0018] In a second aspect, the present invention provides an electrode detection device, comprising: The first module is used to set a detection box in the image of the electrode to be inspected and extract the image features within the detection box. The detection box includes an edge damage detection box for detecting edge and corner damage of the electrode and an electrode tab folding detection box for detecting tab folding defects. The second module is used to input the image features into a preset artificial intelligence model and output the electrode detection results. The preset artificial intelligence model is based on a multi-layer convolutional layer and fully connected layer structure, and optimizes the model parameters through deep learning of different electrode samples.

[0019] Thirdly, the present invention provides an electronic device, including a memory and a processor; The memory is used to store computer programs; The processor is configured to implement the electrode detection method as described in the first aspect when executing the computer program.

[0020] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the electrode detection method as described in the first aspect.

[0021] The beneficial effects of the electrode detection method of the present invention are as follows: by setting edge damage detection boxes and electrode tab folding detection boxes and introducing a preset artificial intelligence model during the electrode image detection process, the automatic identification and classification of electrode appearance defects are realized. By training a large number of electrode defect samples with deep learning, the preset artificial intelligence model can automatically extract image features at different levels and adapt to irregular damage and complex lighting conditions, thereby effectively improving the generalization ability and accuracy of detection. In turn, it can significantly reduce the proportion of over-detection and under-detection caused by false detection of electrode damage and false detection of electrode tab folding. Attached Figure Description

[0022] Figure 1 This is a schematic flowchart of the electrode detection method according to an embodiment of the present invention; Figure 2 This is a logic diagram of the electrode detection method according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the image feature extraction process according to an embodiment of the present invention. Figure 1 ; Figure 4 This is a schematic diagram of the process for extracting edge pixels according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the image feature extraction process according to an embodiment of the present invention. Figure 2 ; Figure 6 This is a schematic diagram of the image feature extraction process according to an embodiment of the present invention. Figure 3 ; Figure 7 This is a system architecture diagram of the electrode detection device according to an embodiment of the present invention; Figure 8 This is a system architecture diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0023] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Although some embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the accompanying drawings and embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0024] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0025] The term "comprising" and its variations as used herein are open-ended, meaning "including but not limited to"; the term "based on" means "at least partially based on"; the term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments"; and the term "optionally" means "optional embodiments". Definitions of other terms will be given in the following description. It should be noted that the concepts of "first," "second," etc., mentioned in this invention are used only to distinguish different devices, modules, or units, and are not intended to limit the order of functions performed by these devices, modules, or units or their interdependencies.

[0026] It should be noted that the terms "one" and "more" used in this invention are illustrative rather than restrictive. Those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0027] The names of the messages or information exchanged between the multiple devices in the embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of these messages or information.

[0028] like Figure 1 As shown, an embodiment of the present invention provides an electrode detection method, comprising: S100: Set a detection box in the image to be inspected of the electrode sheet, and extract the image features within the detection box. The detection box includes an edge damage detection box for detecting edge and corner damage of the electrode sheet and an electrode tab folding detection box for detecting tab folding defects.

[0029] Specifically, in combination Figure 2 As shown, an image of the electrode to be inspected can be acquired using an industrial camera, and multiple detection boxes can be set in the image, such as an edge damage detection box located at the corner of the electrode. Figure 2 Box 1 (as shown) is used to detect edge and corner damage of electrodes, such as the electrode folding detection box located in the electrode tab area. Figure 2 Box 5 (shown) is used to detect tab folding defects, and image features such as grayscale, texture, shape and edge point features are extracted from each detection box.

[0030] S200: Input the image features into a preset artificial intelligence model and output the electrode detection result. The preset artificial intelligence model is based on a multi-layer convolutional layer and fully connected layer structure, and optimizes the model parameters through deep learning of different electrode samples.

[0031] Specifically, image features are input into a pre-defined artificial intelligence model, which is a deep learning model trained on a large number of electrode defect samples. That is, the model parameters are optimized by deep learning on different electrode samples (e.g., defect samples). During the training process, the model parameters are optimized through backpropagation to minimize the prediction error, thereby enabling accurate classification of defects such as irregular damage and electrode tab folding. The pre-defined artificial intelligence model is based on a multi-layer convolutional layer and fully connected layer structure. It can automatically extract hierarchical features of the image through multi-layer convolutional layers and use fully connected layers to perform classification or regression tasks, thereby outputting electrode detection results.

[0032] In this embodiment, by setting edge damage detection boxes and tab folding detection boxes and introducing a preset artificial intelligence model during the electrode image detection process, the automatic identification and classification of electrode appearance defects are realized. By training a large number of electrode defect samples with deep learning, the preset artificial intelligence model can automatically extract image features at different levels, adapt to irregular damage and complex lighting conditions, thereby effectively improving the generalization ability and accuracy of detection, and thus significantly reducing the proportion of over-detection and under-detection caused by false positives of electrode damage and false positives of tab folding.

[0033] Optionally, setting a detection box in the image to be detected of the electrode and extracting image features within the detection box includes: S110: Set the edge damage detection frame in the corner area of ​​the image to be detected.

[0034] Specifically, in combination Figure 3 As shown, when judging the quality of the corners of irregularly shaped electrodes, an edge damage detection box can be set in the corner area of ​​the image to be inspected. The size of the detection box is automatically matched according to the size of the electrode, for example, it can be set as a rectangle with a side length of 10 mm.

[0035] S120: Extract the edge pixels of the image to be detected from the edge damage detection box as the image features.

[0036] Specifically, in combination Figure 3 As shown, edge pixels of the image to be detected are extracted from the edge damage detection box.

[0037] In this optional embodiment, by setting independent edge damage detection boxes in the corner areas of the electrode image, localized and targeted analysis of the detection area can be achieved, avoiding the waste of resources caused by global calculation of the entire image. This allows the algorithm to focus on areas prone to damage, improving detection efficiency. Furthermore, the independent detection boxes can isolate background interference, ensuring more accurate feature extraction in the corner areas, thereby improving the reliability of edge damage recognition.

[0038] Optionally, extracting edge pixels of the image to be detected from the edge damage detection box as the image feature includes: S121: Determine the accurate position of the electrode by grasping the horizontal and vertical edges to establish a local coordinate axis.

[0039] Specifically, in combination Figure 2 and Figure 4 As shown, by grabbing the horizontal edge ( Figure 2 Box 3 shown) and vertical gripping edge ( Figure 2 As shown in box 2), the grayscale changes are scanned along the horizontal and vertical axes to locate the horizontal and vertical boundaries of the electrode. For example, if the grayscale value of a pixel is less than the grayscale value of the pixel to its right and adjacent to it, and the absolute value of the difference between the two is greater than a preset threshold, then the pixel may be the pixel located on the right boundary, which can determine the accurate position of the electrode. The intersection of the two local coordinate axes is defined as the local coordinate origin. The horizontal edge line direction of the horizontal edge recognition is the X-axis direction, and the vertical edge line direction of the vertical edge recognition is the Y-axis direction. A local coordinate system is established for accurate positioning of the corner area.

[0040] S122: Extract the edge pixels based on the local coordinate axes.

[0041] Specifically, in combination Figure 4 As shown, edge pixels are extracted based on local coordinate axes for subsequent geometric analysis. For example, a pixel bandwidth (3 to 5 pixels) can be set around the local coordinate axes as the range of edge search, forming a horizontal search band and a vertical search band, from which edge pixels are extracted.

[0042] In this optional embodiment, the accurate position of the electrode is determined by horizontal and vertical edge grasping operations. After establishing a local coordinate axis, edge pixel extraction is performed, which can achieve precise positioning of the corner area. This effectively eliminates the influence of shooting offset or uneven lighting on the detection results, making the subsequent curve fitting and damage judgment more accurate and repeatable. As a result, the detection results are more stable and reliable, which helps to achieve high-precision judgment of minor damage to the corners of the electrode.

[0043] Optionally, inputting the image features into a preset artificial intelligence model and outputting the electrode detection result includes: The edge pixels are input into the preset artificial intelligence model, and the hierarchical features of the corner region are extracted using a convolutional layer. The damage classification is achieved through a fully connected layer to output the edge corner damage status of the electrode. The edge corner damage status of the electrode includes the damage category and location coordinates.

[0044] Specifically, the extracted edge pixels are input into a preset artificial intelligence model for damage identification. The preset AI model may contain a structure of three convolutional layers and two fully connected layers. The convolutional layers are used to extract morphological features and curvature variation features of the electrode corner regions, and the fully connected layers are used to perform preliminary classification of the damage categories based on these features. In one embodiment, after the preset AI model outputs candidate damage results, the coordinates of the edge pixels can be further fitted. For example, the least squares method can be used to fit the edge pixels to obtain a fitted curve, and the fitted curve is compared with the corresponding local coordinate axes (e.g., ...). Figure 2 As shown by arrow 4), calculate the average distance and angle deviation between the fitted curve and the local coordinate axis. When the deviation exceeds the preset threshold, determine that there is real damage at the edge of the electrode and output the corresponding damage category (e.g., corner chipping, notch or edge breakage) and location coordinates.

[0045] In this optional embodiment, the extracted edge pixels are input into a preset artificial intelligence model for damage identification. Convolutional layers and fully connected layers are used for feature learning and damage classification, which can significantly improve the intelligence level of edge and corner damage identification, thereby achieving accurate classification of irregular damage such as gaps, chipped corners and broken edges of different shapes.

[0046] Optionally, after extracting the edge pixels of the image to be detected from the edge damage detection box, the pole piece detection method further includes: The coordinates of the edge pixels are fitted to determine the fitting curve of the damaged edge. The fitting curve is compared with the corresponding local coordinate axis. The offset or angle deviation between the fitting curve and the local coordinate axis is used to determine whether the edge of the electrode is truly damaged.

[0047] Specifically, the coordinates of edge pixels are fitted, for example, by using the least squares method to fit curves to edge pixels to obtain a fitted curve that represents the shape of the damaged edge. The fitted curve is compared with the corresponding local coordinate axis, and the geometric relationship parameters between the fitted curve and the local coordinate axis are calculated, including but not limited to the average distance offset and angle deviation. When the geometric relationship parameters exceed the preset threshold, it is determined that there is real damage in the corner area of ​​the electrode.

[0048] In one embodiment, the actual damage determination result serves as the basis for inputting the corresponding edge pixels or image features into a preset artificial intelligence model. Only when the damage is determined to be actual, the image features are input into the artificial intelligence model for further damage identification. In another embodiment, the actual damage determination result and the damage identification result output by the preset artificial intelligence model are acquired in parallel, and the final electrode corner damage state is determined based on the consistency of their determinations. In yet another embodiment, after the preset artificial intelligence model outputs the candidate damage result, the actual damage determination result serves as the verification basis to correct or confirm the output result of the preset artificial intelligence model in order to obtain the final electrode detection result.

[0049] In this optional embodiment, by fitting the coordinates of the extracted edge pixels and comparing the fitted damaged edge with the local coordinate axis geometrically, it is possible to determine whether there is real damage at the edge of the electrode from the perspective of geometric consistency. This effectively distinguishes between false edge changes caused by reflection, noise, or local burrs and actual structural damage. Furthermore, the real damage determination result can be used as a pre-filtering condition, parallel verification basis, or post-correction basis for the preset artificial intelligence model. This not only reduces the risk of misjudgment by the preset artificial intelligence model due to abnormal lighting or complex background, but also improves the reliability of detecting irregular edge damage.

[0050] Optionally, setting a detection box in the image to be detected of the electrode and extracting image features within the detection box includes: S130: Set the electrode folding detection frame in the electrode area of ​​the image to be detected, wherein the electrode folding detection frame can be adjusted regionally.

[0051] Specifically, in combination Figure 5 As shown, when judging the damage / folding of the reflective tab foil, a tab folding detection frame is set in the tab area of ​​the image to be detected. The initial position of the tab folding detection frame is determined according to the tab image template. In order to cope with the offset of the tab transmission position, the center position of the tab folding detection frame is adjusted in real time by detecting the coordinate change of the tab reference point to ensure that the detection area of ​​the tab folding detection frame always covers the tab.

[0052] S140: Extract the image features of the image to be detected from the tab folding detection frame.

[0053] Specifically, in combination Figure 5 As shown, image features such as grayscale distribution, brightness peaks, and region contours of the image to be detected are extracted from the electrode folding detection box for subsequent processing.

[0054] Grayscale values ​​are used to represent the brightness or darkness of a single pixel in an image. In a grayscale image, the grayscale value of each pixel represents the brightness of that point, typically ranging from pure black (darkest) to pure white (brightest). A grayscale image is an image that contains only brightness information and does not contain color information.

[0055] In this optional embodiment, by setting a detection frame that can be regionally adjusted in the electrode area, the detection range can be automatically adjusted according to the movement or position change of the electrode, avoiding blind spots caused by shooting offset, ensuring that the electrode detection is always focused on the actual position area, and improving detection stability.

[0056] Optionally, extracting image features of the image to be detected from the electrode folding detection frame includes: S141: Perform grayscale processing on the image within the electrode tab folding detection frame to determine the location of suspected defects.

[0057] Specifically, in combination Figure 2 and Figure 6 As shown, the color image within the tab folding detection frame is converted into a grayscale image. Based on the grayscale threshold, suspected bright areas are initially identified, and the location of suspected defects can be determined.

[0058] S142: Perform grayscale erosion on the image of the region where the suspected defect is located, and extract the image features after morphological transformation of the image after grayscale erosion.

[0059] Specifically, in combination Figure 2 and Figure 6 As shown, grayscale erosion is performed on the image of the area where the suspected defect is located. Figure 2 As shown in Figures 5 and 6), the gray values ​​of discontinuous reflective points are reduced to the surrounding level to weaken pseudo-defects, and then morphological transformation is performed on the eroded image. Figure 2 As shown in Figures 6 to 7), for example, dilation and closure operations are performed to repair the fractured area caused by reflection, and background parts that cannot be flipped are excluded. Finally, the processed regional image features are extracted and input into the preset artificial intelligence model.

[0060] In this optional embodiment, by performing grayscale erosion and morphological transformation operations within the tab detection frame, i.e., grayscale erosion weakens high-brightness noise and morphological operations repair fracture areas, the feature areas become more coherent and realistic, which can effectively reduce false defects caused by reflective metal foil and significantly reduce the misjudgment rate caused by reflection.

[0061] Optionally, inputting the image features into a preset artificial intelligence model and outputting the electrode detection result includes: The image features processed by grayscale erosion and morphological transformation are input into the preset artificial intelligence model. The hierarchical features of the tab region are extracted using convolutional layers, and defect classification is achieved through fully connected layers to output the tab defect status. The tab defect status includes the defect category and location coordinates.

[0062] Specifically, the image features after grayscale erosion and morphological processing are input into a preset artificial intelligence model. The preset artificial intelligence model adopts a convolutional neural network. The input layer size is adjusted according to the detection box size. The convolutional layer extracts the texture, brightness and structural change features of the tabs. The fully connected layer outputs the defect category (such as tab folding, tab breakage, abnormal reflection, etc.) and the corresponding position coordinates.

[0063] In this optional embodiment, the image features after grayscale erosion and morphological transformation are input into an artificial intelligence model for feature extraction and defect classification. This can accurately distinguish different defect types such as tab folding, breakage, and abnormal reflection, achieving precise positioning and confidence assessment, and maintaining high detection accuracy even under complex lighting and reflection interference conditions.

[0064] Optionally, the preset artificial intelligence model adopts a multi-task deep learning model, which includes a shared feature extraction network and multiple sub-models. The shared feature extraction network is used to perform unified hierarchical feature extraction on the input image features, and the multiple sub-models are respectively used to perform different types of defect detection tasks based on the output of the shared feature extraction network.

[0065] Specifically, the preset artificial intelligence model adopts a multi-task deep learning model architecture to uniformly detect various defect types in electrode images. The multi-task deep learning model includes a shared feature extraction network and multiple sub-models. The shared feature extraction network is set at the front end of the model and is used to uniformly encode the input electrode image to be detected or the image features extracted by the detection box. The shared feature extraction network may include multiple convolutional layers and non-linear activation layers. It extracts edge information, texture features and morphological change features in the electrode image through layer-by-layer convolution operations, thereby forming a hierarchical feature representation. The output features of the shared feature extraction network are simultaneously input to multiple sub-models. Each sub-model performs different types of defect detection tasks based on the shared features, thereby realizing parallel detection of electrode corner damage and electrode tab folding defects.

[0066] In this optional embodiment, a multi-task deep learning model, including a shared feature extraction network and multiple sub-models, is used in the electrode detection method to perform unified hierarchical feature extraction on the input image features, and different types of defect detection tasks are performed based on the shared features. This enables different detection tasks to share at the underlying feature level, reduces computational redundancy caused by repeated feature extraction, and improves the overall detection efficiency.

[0067] Optionally, the sub-model includes a first sub-model for detecting electrode corner damage. The first sub-model analyzes the edge pixel features of the electrode corner region based on the feature information output by the shared feature extraction network, classifies the edge pixel features for damage, and outputs the electrode corner damage status.

[0068] Specifically, the first sub-model in the multi-task deep learning model is used to detect edge and corner damage defects of the electrode sheet. The first sub-model receives feature output from the shared feature extraction network and further analyzes the features according to the detection requirements of the electrode sheet edge and corner region. The first sub-model takes the extracted edge pixel features as input. The edge pixel features can be obtained by edge extraction algorithms from the pixels within the edge damage detection box. The first sub-model extracts morphological and geometric change features of the edge pixel features through convolution operations, and performs classification calculations on the extracted features through fully connected layers, thereby determining whether there is damage at the electrode sheet edge and corner. The output of the first sub-model is the electrode sheet edge and corner damage status, which includes damage category information and corresponding spatial coordinates, used to characterize the damage type and its specific location in the electrode sheet.

[0069] In this optional embodiment, a first sub-model based on a shared feature extraction network is set up to perform targeted analysis on the edge pixel features of the electrode corner region, and the edge pixel features are classified for damage, thereby achieving accurate identification of electrode corner damage defects. This method can fully utilize the geometric shape and structural change features reflected by the edge pixels, improve the accuracy of identifying defects such as irregular corner damage, chipped corners, or gaps, while avoiding invalid analysis of non-corner regions, thereby reducing the false detection rate and improving detection reliability.

[0070] Optionally, the sub-model includes a second sub-model for detecting tab folding defects. The second sub-model analyzes the image features of the tab region based on the feature information output by the shared feature extraction network, classifies the image features of the tab region as defects, and outputs the tab defect status.

[0071] Specifically, the second sub-model in the multi-task deep learning model is used to detect tab folding defects. The second sub-model also receives feature information output by the shared feature extraction network and performs feature analysis on the structural characteristics of the tab region. The image features of the tab region can first undergo preprocessing operations such as grayscale processing, grayscale erosion, and morphological transformation to reduce reflective interference and highlight suspected defect areas. The preprocessed tab region image features are input to the second sub-model. The second sub-model extracts local structural features and morphological change features of the tab region through convolutional layers and performs defect classification on the features through fully connected layers. The second sub-model outputs the tab defect status, which includes defect category information such as whether the tab has folded or broken, as well as the corresponding spatial coordinates.

[0072] In this optional embodiment, a second sub-model for detecting tab folding defects is set up, and the image features of the tab region are analyzed based on the feature information output by the shared feature extraction network, thereby achieving specialized detection of tab folding and related defects. This detection method can extract discriminative structural and morphological features that are easily reflected and have complex morphological changes in the tab region, thereby improving the recognition accuracy of tab folding defects, reducing misjudgments caused by changes in illumination or local noise, and enhancing the stability of tab defect detection.

[0073] Optionally, the multi-task deep learning model further includes a task collaborative judgment module, which is used to perform correlation analysis on the electrode edge damage state or electrode tab defect state output by the sub-model. Based on the correlation of different defect types in spatial location, morphological characteristics or occurrence probability, the detection results of the sub-model are comprehensively corrected or the confidence level is adjusted to output the final electrode detection result.

[0074] Specifically, the multi-task deep learning model also includes a task collaborative judgment module, which is used to collaboratively analyze the detection results of the first sub-model and the second sub-model. The task collaborative judgment module receives the electrode corner damage status output by the first sub-model and the electrode tab defect status output by the second sub-model, and performs correlation analysis on the two types of detection results. Considering the spatial / morphological correlation between corner damage and electrode tab folding in the same electrode, the task collaborative judgment module can make a comprehensive judgment on the detection results based on the spatial distribution relationship, morphological feature correlation, or historical occurrence probability of different defect types in the electrode. For example, when there is a preset correlation between corner damage and electrode tab folding in spatial location, the task collaborative judgment module can adjust the confidence level of the corresponding detection result. In one embodiment, the task collaborative judgment module corrects the misjudgment results that may be generated by a single sub-model by weighting the confidence level information output by each sub-model. In another embodiment, the task collaborative judgment module performs consistency verification on the detection results based on the correlation analysis results to output the final electrode detection result.

[0075] In this optional embodiment, by introducing a task collaborative judgment module into the multi-task deep learning model, the electrode corner damage state and electrode tab defect state output by the first sub-model and the second sub-model are correlated and analyzed. Based on the correlation of different defect types in spatial location, morphological characteristics or occurrence probability, the detection results are comprehensively corrected or the confidence level is adjusted. This can effectively reduce the risk of misjudgment caused by independent judgment of a single sub-model, and make the final output electrode detection results more consistent with the actual defect distribution characteristics, thereby further improving the overall accuracy and reliability of multi-defect detection.

[0076] like Figure 7 As shown, an embodiment of the present invention provides an electrode detection device 700, comprising: The first module 710 is used to set a detection box in the image to be inspected of the electrode sheet and extract image features within the detection box. The detection box includes an edge damage detection box for detecting edge and corner damage of the electrode sheet and an electrode tab folding detection box for detecting tab folding defects. The second module 720 is used to input the image features into a preset artificial intelligence model and output the electrode detection results. The preset artificial intelligence model is based on a multi-layer convolutional layer and fully connected layer structure, and optimizes the model parameters through deep learning of different electrode samples.

[0077] like Figure 8 As shown, an electronic device 800 provided in this embodiment of the invention includes a memory 820 and a processor 810; the memory 820 is used to store a computer program; the processor 810 is used to implement the electrode detection method as described above when the computer program is executed.

[0078] Alternatively, an electronic device 800 includes a memory 820 and a processor 810 coupled to the memory 820; the memory 820 is configured to store a computer program; and the processor 810 is configured to perform the following operations when the computer program is executed: A detection box is set in the image of the electrode to be inspected, and the image features within the detection box are extracted. The detection box includes an edge damage detection box for detecting edge and corner damage of the electrode and an electrode tab folding detection box for detecting tab folding defects. The image features are input into a preset artificial intelligence model, and the pole detection results are output. The preset artificial intelligence model is based on a multi-layer convolutional layer and fully connected layer structure, and the model parameters are optimized through deep learning of different pole samples.

[0079] This invention provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the electrode detection method described above.

[0080] Alternatively, a non-volatile computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform the following operations: A detection box is set in the image of the electrode to be inspected, and the image features within the detection box are extracted. The detection box includes an edge damage detection box for detecting edge and corner damage of the electrode and an electrode tab folding detection box for detecting tab folding defects. The image features are input into a preset artificial intelligence model, and the pole detection results are output. The preset artificial intelligence model is based on a multi-layer convolutional layer and fully connected layer structure, and the model parameters are optimized through deep learning of different pole samples.

[0081] The present invention will now be described an electronic device 800 that can serve as a server or client of the present invention, which is an example of a hardware device that can be applied to various aspects of the present invention. Electronic device 800 is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic device 800 can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0082] Electronic device 800 includes a computing unit that can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) or a computer program loaded from a storage unit into random access memory (RAM). The RAM may also store various programs and data required for device operation. The computing unit, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.

[0083] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc. In this application, the units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of the present invention according to actual needs. Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated units can be implemented in hardware or as software functional units.

[0084] While the present invention has been disclosed above, its scope of protection is not limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention, and all such changes and modifications will fall within the scope of protection of the present invention.

Claims

1. A method for detecting electrode sheets, characterized in that, include: A detection box is set in the image of the electrode to be inspected, and the image features within the detection box are extracted. The detection box includes an edge damage detection box for detecting edge and corner damage of the electrode and an electrode tab folding detection box for detecting tab folding defects. The image features are input into a preset artificial intelligence model, and the pole detection results are output. The preset artificial intelligence model is based on a multi-layer convolutional layer and fully connected layer structure, and the model parameters are optimized through deep learning of different pole samples.

2. The electrode detection method according to claim 1, characterized in that, The step of setting a detection box in the image to be detected of the electrode and extracting image features within the detection box includes: The edge damage detection frame is set in the corner area of ​​the image to be detected; The edge pixels of the image to be detected are extracted from the edge damage detection box as the image features.

3. The electrode detection method according to claim 2, characterized in that, The step of extracting edge pixels of the image to be detected from the edge damage detection box as the image feature includes: The precise position of the electrode is determined by grasping the horizontal and vertical edges to establish a local coordinate axis; The edge pixels are extracted based on the local coordinate axes.

4. The electrode detection method according to claim 2, characterized in that, The step of inputting the image features into a preset artificial intelligence model and outputting the electrode detection result includes: The edge pixels are input into the preset artificial intelligence model, and the hierarchical features of the corner region are extracted using a convolutional layer. The damage classification is achieved through a fully connected layer to output the edge corner damage status of the electrode. The edge corner damage status of the electrode includes the damage category and location coordinates.

5. The electrode detection method according to claim 2, characterized in that, After extracting the edge pixels of the image to be detected from the edge damage detection box, the pole piece detection method further includes: The coordinates of the edge pixels are fitted to determine the fitting curve of the damaged edge. The fitting curve is compared with the corresponding local coordinate axis. The offset or angle deviation between the fitting curve and the local coordinate axis is used to determine whether the edge of the electrode is truly damaged.

6. The electrode detection method according to claim 1, characterized in that, The step of setting a detection box in the image to be detected of the electrode and extracting image features within the detection box includes: A tab folding detection frame is set in the tab area of ​​the image to be detected, wherein the tab folding detection frame can be regionally adjusted. Image features of the image to be detected are extracted from the electrode folding detection frame.

7. The electrode detection method according to claim 6, characterized in that, The step of extracting image features of the image to be detected from the electrode flapping detection frame includes: The image within the electrode flap folding detection frame is processed in grayscale to determine the location of suspected defects; The image of the region where the suspected defect is located is subjected to grayscale erosion, and the image features are extracted after morphological transformation of the grayscale eroded image.

8. The electrode detection method according to claim 7, characterized in that, The step of inputting the image features into a preset artificial intelligence model and outputting the electrode detection result includes: The image features processed by grayscale erosion and morphological transformation are input into the preset artificial intelligence model. The hierarchical features of the tab region are extracted using convolutional layers, and defect classification is achieved through fully connected layers to output the tab defect status. The tab defect status includes defect category and location coordinates.

9. The electrode detection method according to claim 1, characterized in that, The preset artificial intelligence model adopts a multi-task deep learning model, which includes a shared feature extraction network and multiple sub-models. The shared feature extraction network is used to perform unified hierarchical feature extraction on the input image features, and the multiple sub-models are used to perform different types of defect detection tasks based on the output of the shared feature extraction network.

10. The electrode detection method according to claim 9, characterized in that, The sub-model includes a first sub-model for detecting electrode corner damage. The first sub-model analyzes the edge pixel features of the electrode corner region based on the feature information output by the shared feature extraction network, classifies the edge pixel features for damage, and outputs the electrode corner damage status.

11. The electrode detection method according to claim 9, characterized in that, The sub-model includes a second sub-model for detecting tab folding defects. The second sub-model analyzes the image features of the tab region based on the feature information output by the shared feature extraction network, classifies the image features of the tab region for defects, and outputs the tab defect status.

12. The electrode detection method according to claim 9, characterized in that, The multi-task deep learning model also includes a task collaborative judgment module, which is used to perform correlation analysis on the electrode edge damage state or electrode tab defect state output by the sub-model. Based on the correlation of different defect types in spatial location, morphological characteristics or occurrence probability, the detection results of the sub-model are comprehensively corrected or the confidence level is adjusted to output the final electrode detection result.

13. An electrode detection device, characterized in that, include: The first module is used to set a detection box in the image of the electrode to be inspected and extract the image features within the detection box. The detection box includes an edge damage detection box for detecting edge and corner damage of the electrode and an electrode tab folding detection box for detecting tab folding defects. The second module is used to input the image features into a preset artificial intelligence model and output the electrode detection results. The preset artificial intelligence model is based on a multi-layer convolutional layer and fully connected layer structure, and optimizes the model parameters through deep learning of different electrode samples.

14. An electronic device, characterized in that, Including memory and processor; The memory is used to store computer programs; The processor is configured to implement the electrode detection method as described in any one of claims 1 to 12 when executing the computer program.

15. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the electrode detection method as described in any one of claims 1 to 12.

Images