Defect detection method and system

Abstract

The present application belongs to the technical field of defect detection. Disclosed are a defect detection method and system. The present application involves: acquiring battery cell images which are collected at a plurality of angles, wherein each battery cell image comprises a tab portion of a battery cell; by means of a target detection model, performing tab defect detection on said battery cell images, so as to obtain a tab detection result, wherein the target detection model is used for performing tab positioning and tab defect identification; and completing defect detection by means of the tab detection result. By means of a target detection model, tab defect detection is performed on battery cell images which are collected at a plurality of angles, such that a defect in a tab portion can be accurately detected; and compared with manual identification for a tab defect, the present application improves the efficiency and precision of tab defect detection, and increases the rate of identification for a minor tab defect.

Description

Defect detection method and system

[0001] Priority information

[0002] This application claims priority to Chinese patent application No. 202311679792.5 filed on December 7, 2023, the entire contents of which are incorporated by reference into this application. Technical Field

[0003] The present application relates to the field of defect detection technology, and in particular to a defect detection method and system. Background Art

[0004] Currently, during the battery cell production process, multiple handling operations occur during the adapter welding, top cover welding, and cell assembly process, which can cause cracking and wrinkling in the battery cell tabs. Cracks or severe wrinkling in the tabs will block the flow of current in the battery cell. In severe cases, it can even cause the battery cell to overheat, leading to serious safety issues such as battery cell fires. The tab cracking problem has always been a difficult and painful point that needs to be solved in battery cell production. Therefore, solving the tab cracking problem during battery cell production can greatly reduce the safety risk of the finished battery cell and improve product quality.

[0005] Currently, tab defect detection uses manual visual inspection or charge-discharge cycle testing of each battery cell. This method cannot detect subtle tab cracks and requires charging and discharging each battery cell, resulting in low detection efficiency and accuracy.

[0006] Application Contents

[0007] The main purpose of this application is to provide a defect detection method and system, aiming to solve the technical problems of low detection accuracy and efficiency in the existing technology.

[0008] In a first aspect, an embodiment of the present application provides a defect detection method, which includes the following steps:

[0009] Acquire battery cell images collected from multiple angles, wherein the battery cell images include tabs of the battery cell;

[0010] Performing tab defect detection on battery cell images collected from multiple angles using a target detection model to obtain tab detection results. The target detection model is used to locate and identify tab defects.

[0011] Defect detection is completed through the tab detection results.

[0012] This embodiment proposes a defect detection method, which includes: obtaining battery cell images collected from multiple angles, wherein the battery cell images include the tab portion of the battery cell; performing tab defect detection on the battery cell images collected from multiple angles using a target detection model to obtain a tab detection result, wherein the target detection model is used to locate the tab and identify tab defects; completing defect detection using the tab detection result, and performing tab defect detection on the battery cell images collected from multiple angles using the target detection model to accurately detect defects in the tab portion. Compared with manual identification of tab defects, the efficiency and accuracy of tab defect detection and the recognition rate of minor tab defects are improved.

[0013] In some embodiments, the object detection model includes an input layer, a feature extraction layer, a feature fusion layer, and an output layer;

[0014] The method of performing tab defect detection on battery cell images collected from multiple angles using a target detection model to obtain tab detection results includes:

[0015] The battery cell images collected from multiple angles are enhanced through the input layer of the target detection model to obtain enhanced battery cell images.

[0016] Performing feature extraction on the battery cell enhanced image through a feature extraction layer in the target detection model to obtain extracted features;

[0017] Performing feature fusion on the extracted features through a feature fusion layer in the target detection model to obtain fused features;

[0018] The output layer in the target detection model performs tab position detection or tab defect detection on the fused features to obtain a tab detection result.

[0019] In this embodiment, corresponding processing is performed through each structural layer in the target detection model, so that image features can be extracted quickly and accurately, and tab position detection or tab defect detection can be performed based on the image features, which can meet the requirements of quickly completing real-time tab defect detection without affecting the production of battery cells.

[0020] In some embodiments, the object detection model includes a first object detection model, which is trained using battery cell sample images labeled with tab regions and tab misalignment regions;

[0021] The method of performing tab defect detection on battery cell images collected from multiple angles using a target detection model to obtain tab detection results includes:

[0022] Performing image enhancement on battery cell images collected from multiple angles through the input layer of the first target detection model to obtain battery cell enhanced images;

[0023] Performing feature extraction on the battery cell enhanced image through the feature extraction layer in the first object detection model to obtain a first tab extraction feature;

[0024] Performing feature fusion on the first tab extraction features through a feature fusion layer in the first object detection model to obtain a first tab fusion feature;

[0025] Performing tab location on the first tab fusion feature through the output layer in the first object detection model to obtain a tab detection area;

[0026] The tab detection result is obtained through the tab detection area.

[0027] In the technical solution of the embodiment of the present application, the tab area in the battery cell image is detected by the first target detection model, the tab position can be determined first, and the tab can be roughly positioned, thereby improving the accuracy of tab defect detection.

[0028] In some embodiments, the defect detection method further includes:

[0029] The battery cell image is cropped according to the tab detection area to obtain a tab image.

[0030] In this embodiment, after the tab detection area is identified, the battery cell image is cropped into the tab image, so that the tab defects can be accurately identified directly through the tab image.

[0031] In some embodiments, the object detection model further includes a second object detection model, wherein the second object detection model is trained using tab sample images marked with tab defects;

[0032] The step of obtaining a tab detection result through the tab detection area includes:

[0033] Inputting the tab image into the second object detection model;

[0034] Performing image enhancement on the tab image through the input layer in the second object detection model to obtain a tab enhanced image;

[0035] Performing feature extraction on the tab enhanced image through a feature extraction layer in the second object detection model to obtain a second tab extraction feature;

[0036] Performing feature fusion on the second tab extraction features through a feature fusion layer in the second object detection model to obtain a second tab fusion feature;

[0037] Performing tab defect detection on the second tab fusion feature through the output layer in the second object detection model to obtain a tab defect category;

[0038] The tab detection result is obtained according to the tab defect category.

[0039] In the technical solution of the embodiment of the present application, defects in the tab image are detected through a second target detection model, which can quickly detect the tab defect category, solve the tab misalignment problem that occurs in the battery cell production process, and improve the recognition rate of small defects and minor defects.

[0040] In some embodiments, obtaining the tab detection result according to the tab defect category includes:

[0041] Obtaining the defect length corresponding to the tab defect category;

[0042] comparing the defect length with a preset length threshold;

[0043] When the defect length is greater than the preset length threshold, the tab inspection result is confirmed as a defect in the tab.

[0044] In this embodiment, the defect length corresponding to the tab defect category is obtained and the defect length is compared with a standard length threshold, thereby quickly determining whether the tab length has a defect.

[0045] In some embodiments, the defect detection method further includes:

[0046] When the defect length is less than or equal to the preset length threshold, the tab inspection result is confirmed to be that the tab has no defects.

[0047] In this embodiment, by comparing the defect length with a standard length threshold, when the defect length is less than or equal to the standard length threshold, it is quickly determined that there is no defect in the tab length.

[0048] In some embodiments, obtaining the tab detection result according to the tab defect category includes:

[0049] Obtaining the defect area corresponding to the tab defect category;

[0050] comparing the defect area with a preset area threshold;

[0051] When the defect area is greater than the preset area threshold, the tab inspection result is confirmed as a defect in the tab.

[0052] In this embodiment, the defect area corresponding to the tab defect category is obtained and compared with the standard area threshold, thereby quickly determining whether there is a defect on the tab area.

[0053] In some embodiments, the defect detection method further includes:

[0054] When the defect area is less than or equal to the preset area threshold, the tab inspection result is confirmed to be that the tab has no defects.

[0055] In the technical solution of the embodiment of the present application, by comparing the defect area with the standard area threshold, when the defect area is less than or equal to the standard area threshold, it is quickly determined that there is no defect in the tab area.

[0056] In some embodiments, obtaining the tab detection result through the tab detection area includes:

[0057] Detecting whether there is tab misalignment in the tab detection area;

[0058] When there is tab misalignment in the tab detection area, the tab detection result is determined to be tab misalignment.

[0059] In this embodiment, the tab is detected to determine whether it is misaligned in the tab detection area obtained by the first target detection model, thereby enabling detection of tab misalignment defects and improving the tab detection range.

[0060] In some embodiments, the defect detection method further includes:

[0061] The tab misalignment region in the tab image is cropped to obtain a cropped tab image.

[0062] In this embodiment, when there is tab misalignment in the tab image, the tab misalignment region in the tab image is cropped to obtain a cropped tab image, thereby improving the accuracy of subsequent tab defect detection.

[0063] In some embodiments, before performing tab defect detection on battery cell images collected from multiple angles using a target detection model and obtaining tab detection results, the method further includes:

[0064] Acquire a battery cell sample image, wherein the battery cell sample image is marked with a tab region and a tab misalignment region;

[0065] An initial target detection network is trained using the battery cell sample images to obtain a first target detection model.

[0066] In this embodiment, a battery cell sample image is obtained, and the battery cell sample image is annotated with the tab area and the tab misalignment area, so that the initial target detection network is trained through the battery cell sample image of the tab area and the tab misalignment area, so that the position of the tab part and the misalignment area obtained for the input battery cell image are close to the annotated real information, thereby achieving the purpose of detecting the tab position and the tab misalignment area.

[0067] In some embodiments, before performing tab defect detection on battery cell images collected from multiple angles using a target detection model and obtaining tab detection results, the method further includes:

[0068] Acquire a tab sample image, wherein the tab sample image is marked with a tab defect;

[0069] The initial target detection network is trained using the tab sample images to obtain a second target detection model.

[0070] In the technical solution of the embodiment of the present application, by obtaining a sample image of the tab, the sample image of the tab is marked with tab defects, and the initial target detection network is trained using the sample image of the tab. Since a large amount of information will be lost during the scaling process of the image, the second target detection model is obtained by training using the sample image of the tab to improve the model training accuracy.

[0071] In some embodiments, completing defect detection based on the tab inspection result includes:

[0072] When the tab detection result indicates that the tab has a defect, output the tab defect category and the tab defect location;

[0073] It is determined that the battery cell corresponding to the tab does not meet the production requirements, and the battery cell is discarded.

[0074] In this embodiment, when the tab detection result shows that the tab is defective, the specific tab defect category and tab defect location can be output, so that the battery cell to which the defective tab belongs is judged as not meeting production requirements and is removed, thereby realizing intelligent control of the production line and effectively ensuring the production line rhythm.

[0075] In a second aspect, an embodiment of the present application further provides a defect detection system, the defect detection system comprising:

[0076] Image acquisition module, used to capture images of battery cells including the tabs at multiple angles;

[0077] The host computer is used to confirm whether there are defects in the tab based on the battery cell images collected from multiple angles.

[0078] In the technical solution of the embodiment of the present application, the image acquisition module is used to capture images of the battery cell including the tab area at multiple angles, so that a complete battery cell image can be obtained, the accuracy of detection can be improved, and the host computer can be used to confirm whether the tab has defects, thereby improving the efficiency and accuracy of tab defect detection and the recognition rate of minor tab defects.

[0079] In some embodiments, the image acquisition modules include two, each image acquisition module includes at least two image acquisition devices, and the two image acquisition modules are respectively located at the end of the anode ear and the end of the cathode ear;

[0080] The image acquisition module is used to perform multi-angle image acquisition on the battery cell image including the tab portion, obtain the multi-angle acquired battery cell image and send it to the host computer.

[0081] In the technical solution of the embodiment of the present application, two image acquisition modules are used to respectively acquire images of battery cells including anode ears and images of battery cells including cathode ears, so as to perform ear defect detection separately, thereby avoiding the mixing of anode ear images and cathode ear images to affect the detection results.

[0082] In some embodiments, the anode ears include two, the cathode ears include two; the image acquisition module includes a first image acquisition module and a second image acquisition module;

[0083] The two image acquisition devices in the first image acquisition module are respectively located at the ends of the two anode ears, and the two image acquisition devices in the second image acquisition module are respectively located at the ends of the two cathode ears.

[0084] In the technical solution of the embodiment of the present application, multiple image acquisition devices are used to capture the anode ear or cathode ear from multiple angles, so as to fully obtain the surface of the electrode ear, obtain a complete electrode ear image, and improve the accuracy of detection.

[0085] In some embodiments, the system further comprises at least one light source, wherein the at least one light source and the image acquisition module are disposed on two sides of the battery cell opposite to each other;

[0086] At least one of the light sources is used to provide light source when the image acquisition module acquires an image of a battery cell including a tab portion.

[0087] In the technical solution of the embodiment of the present application, a light source is further provided in the system to provide fill light when the image acquisition device acquires the battery cell image, so that a clear image of the battery cell including the tab portion can be captured, thereby improving the shooting effect.

[0088] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0089] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. 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 the structures shown in these drawings without paying any creative work.

[0090] FIG1 is a flow chart of an embodiment of a defect detection method proposed in an embodiment of the present application;

[0091] FIG2 is another flow chart of an embodiment of a defect detection method proposed in an embodiment of the present application;

[0092] FIG3 is another flow chart of an embodiment of a defect detection method proposed in an embodiment of the present application;

[0093] FIG4 is another flow chart of an embodiment of a defect detection method proposed in an embodiment of the present application;

[0094] FIG5 is a schematic diagram of the overall process of defect detection in an embodiment of the defect detection method proposed in an embodiment of the present application;

[0095] FIG6 is a schematic structural diagram of an embodiment of defect detection proposed in an embodiment of the present application;

[0096] FIG7 is another structural diagram of an embodiment of defect detection proposed in an embodiment of the present application.

[0097] The reference numerals in the specific embodiments are as follows: defect detection system 1; image acquisition module 10, image acquisition device 101; host computer 20, light source 30; electrode tab 2; anode tab 21, cathode tab 22. DETAILED DESCRIPTION

[0098] The following embodiments of the technical solution of the present application will be described in detail with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present application and are therefore only examples and are not intended to limit the scope of protection of the present application.

[0099] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application belongs; the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned figure descriptions are intended to cover non-exclusive inclusions.

[0100] In the description of the embodiments of this application, the technical terms "first" and "second" are used only to distinguish different objects and should not be understood to indicate or imply relative importance or implicitly specify the quantity, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, the meaning of "plurality" is more than two, unless otherwise clearly and specifically defined.

[0101] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0102] In the description of the embodiments of this application, the term "and / or" is simply a description of the association relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent the following three situations: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this document generally indicates that the associated objects are in an "or" relationship.

[0103] In the description of the embodiments of the present application, the term "multiple" refers to more than two (including two). Similarly, "multiple groups" refers to more than two groups (including two groups), and "multiple pieces" refers to more than two pieces (including two pieces).

[0104] In the description of the embodiments of the present application, the technical terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing the embodiments of the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the embodiments of the present application.

[0105] In the description of the embodiments of the present application, unless otherwise expressly specified or limited, technical terms such as "installed," "connected," "connected," and "fixed" should be understood in a broad sense. For example, they can refer to fixed connections, detachable connections, or integration; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; internal connections between two components or interactions between two components. Those skilled in the art can understand the specific meanings of the above terms in the embodiments of the present application based on specific circumstances.

[0106] Currently, the most common method for detecting tab defects is manual visual inspection. This requires a person to manually inspect the tabs of each battery cell during the battery cell production process to check for defects such as tab cracking. However, this method is slow, inefficient, and has a low detection rate. It requires significant manpower and cannot guarantee 100% detection of defective tab fractures. Alternatively, the battery is subjected to charge and discharge cycles, and the cycle data is recorded. The average charge or discharge voltage of the battery under test during each cycle is calculated to identify abnormal cycles. Based on the charge and discharge curves of the abnormal cycles, if any data point deviates from the charge and discharge curve by a value greater than a first preset threshold, a tab fracture is determined. This method of continuously charging and discharging the battery to detect defects can only detect larger defects and cannot detect subtle tab defects. Furthermore, the detection process requires charging and discharging each battery, which is complex and inefficient.

[0107] Based on the above considerations, in order to improve the detection accuracy and efficiency of tab defect detection, after in-depth research, a defect detection method and system were proposed, which are suitable for various scenarios of defect detection in the battery production process, such as tab defect detection during battery cell adapter welding, top cover welding or core assembly. Due to the multiple handling operations in the above process, the tabs of the battery cells may cause cracking, wrinkling and other problems, thereby blocking the flow of current in the battery cells. In severe cases, it may even cause excessive heating of the battery cells, leading to serious safety problems such as battery cell fire.

[0108] This application builds a target detection model in advance based on a deep learning algorithm, and uses an industrial camera to shoot the tab part of the battery cell at multiple angles while the battery cell is being transported or when the gripper is grabbing the battery cell, thereby obtaining multi-angle battery cell images including the tab part. The target detection model is used to locate the tab and detect defects in the multi-angle battery cell images, which can quickly and accurately complete the inspection and improve the inspection accuracy and efficiency.

[0109] In order to better understand the embodiments of the present application, the defect detection method and system provided according to the embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0110] The executor of the defect detection method provided in the embodiment of the present application may be a defect detection system. The defect detection system may be used to detect various defects on the tabs, and may also be used to detect defects on other objects, which is not limited in this embodiment.

[0111] In some embodiments of the present application, referring to FIG. 1 , an embodiment of the present application proposes a defect detection method. The defect detection method may include:

[0112] Step S10: Acquire battery cell images collected from multiple angles, where the battery cell images include the tabs of the battery cells.

[0113] It should be noted that the battery cell in this embodiment is the battery core. Due to the production process, the tab adapter of the battery cell may have been welded, and the tab surface has become an arc shape. The tab surface cannot be fully obtained from one angle. Therefore, in order to ensure the accuracy of the detection, battery cell images collected from multiple angles can be obtained.

[0114] In a specific implementation, multiple image acquisition devices may be used to acquire battery cell images at multiple angles, thereby obtaining a complete battery cell image.

[0115] It should be noted that the battery cell image includes the tab portion of the battery cell and may also include other portions. By collecting battery cell images in a larger range, a complete image of the tab portion can be obtained.

[0116] Step S20: performing tab defect detection on the battery cell images collected from multiple angles through a target detection model to obtain tab detection results. The target detection model is used to perform tab positioning and tab defect identification.

[0117] It should be noted that the target detection model can be established in advance by collecting multiple image samples and labeling the image samples, thereby training and generating the target detection model.

[0118] As an implementation method, the target detection model can be used to perform tab positioning and tab defect identification. The target detection model can also only perform tab defect identification or tab positioning.

[0119] In a specific implementation, the target detection model is mainly used for target detection and can be obtained by training the YOLOv5 network. The target detection model is used to detect tab defects in battery cell images to obtain tab detection results.

[0120] It should be noted that the tab inspection result may be that the tab has defects or that the tab does not have defects.

[0121] For example, the tab defect detection result may be the category of the tab defect or information such as the length, width, and area of ​​the tab defect.

[0122] Step S30: completing defect detection based on the tab detection results.

[0123] It should be understood that defect detection can be completed through the tab inspection results. For example, if the tab inspection result shows that the tab is defective, the specific defect category and defect location can be output, so that secondary inspection can be performed based on the defect category and defect location, or the battery cells corresponding to the defective tabs can be removed from the production line and the battery cells can be disposed of as waste.

[0124] The defect detection method proposed in this embodiment obtains battery cell images collected from multiple angles, wherein the battery cell images include the tab portion of the battery cell; performs tab defect detection on the battery cell images collected from multiple angles through a target detection model to obtain a tab detection result, and the target detection model is used to perform tab positioning and tab defect identification; defect detection is completed using the tab detection result, and tab defect detection is performed on the battery cell images collected from multiple angles through the target detection model, so that defects in the tab portion can be accurately detected. Compared with manual identification of tab defects, the efficiency and accuracy of tab defect detection and the recognition rate of minor tab defects are improved.

[0125] In some embodiments, the object detection model includes an input layer, a feature extraction layer, a feature fusion layer, and an output layer.

[0126] It should be noted that the target detection model is obtained by training the yolov5 initial target detection network. The yolov5 network structure mainly consists of four parts: input end, skeleton network (Backone), Neck and Prediction. Therefore, the trained target detection model mainly consists of input layer, feature extraction layer (Backone), feature fusion layer (Neck) and output layer (Head (Prediction)).

[0127] The input layer is used to perform image enhancement on the input image, such as image cropping and splicing, random scaling, random cropping, random arrangement, affine transformation and other operations.

[0128] The feature extraction layer is used to extract features. The feature extraction layer includes a convolutional neural network that aggregates and forms image features at different image granularities, thereby performing feature extraction and obtaining image features.

[0129] The feature fusion layer is used to fuse features, including a series of network layers that mix and combine image features, perform feature fusion, and pass the fused features to the output layer.

[0130] The output layer is used to perform related detection or prediction, such as predicting image features, generating bounding boxes, and predicting categories.

[0131] In some embodiments, referring to FIG. 2 , the step of performing tab defect detection on battery cell images collected from multiple angles using an object detection model to obtain tab detection results includes:

[0132] Step S201: performing image enhancement on the battery cell images collected from multiple angles through the input layer in the target detection model to obtain a battery cell enhanced image.

[0133] It should be understood that the battery cell images collected from multiple angles can be enhanced through the input layer in the target detection model, such as by cropping and splicing, so as to improve the quality and clarity of the image and obtain a battery cell enhanced image.

[0134] It should be noted that image enhancement also includes pre-processing of battery cell images collected from multiple angles, including denoising, removing artifacts, and adjusting image brightness and contrast, thereby reducing the impact of noise and artifacts.

[0135] Step S202: extracting features from the enhanced image of the battery cell through the feature extraction layer in the target detection model to obtain extracted features.

[0136] In this embodiment, the feature extraction layer in the target detection model can perform feature extraction on the battery cell enhanced image, thereby extracting tab features or tab defect features, etc.

[0137] The feature extraction layer in the target detection model includes a convolutional neural network, which extracts features from the battery cell enhanced image through the convolution kernel in the convolutional neural network to obtain extracted features.

[0138] In a specific implementation, multiple convolution blocks can be combined together, and each convolution block can contain different convolution kernel sizes and strides to capture features of different scales and obtain multi-scale extracted features.

[0139] Step S203: performing feature fusion on the extracted features through the feature fusion layer in the target detection model to obtain fused features.

[0140] It should be noted that the feature fusion layer in the target detection model is used to fuse the multi-scale extracted features to obtain fused features.

[0141] In practice, low-level features in multi-scale feature extraction have higher resolution and contain more positional and detailed information. However, due to fewer convolutions, they are less semantic and contain more noise. High-level features have stronger semantic information but are lower in resolution and have poorer perception of detail. Therefore, features extracted at different scales are fused to improve performance.

[0142] Step S204: performing tab position detection or tab defect detection on the fused features through the output layer in the target detection model to obtain tab detection results.

[0143] It should be understood that the output layer in the target detection model is the prediction layer, which predicts the fused features output by the feature fusion layer to generate a bounding box and a predicted category.

[0144] In a specific implementation, the fused features are predicted through the output layer to obtain a bounding box for the tab location and a predicted target category of the tab. This may also include a bounding box for the tab location on the battery cell and a predicted target category of the tab, with the tab location and predicted target category being the tab detection result.

[0145] In a specific implementation, the fused features are predicted through the output layer to obtain the bounding boxes of various tab defects and the predicted specific tab defect categories, such as tab cracks and tab wrinkles. The location of the tab defect and the specific category of the tab defect are used as the tab detection result.

[0146] In this embodiment, corresponding processing is performed through each structural layer in the target detection model, so that image features can be extracted quickly and accurately, and tab position detection or tab defect detection can be performed based on the image features, which can meet the requirements of quickly completing real-time tab defect detection without affecting the production of battery cells.

[0147] In some embodiments, in order to improve the accuracy of tab defect detection, the tab of the captured battery cell image can be first located using a target detection model, and then the tab-located image can be used for tab defect detection, thereby improving the accuracy of tab defect detection.

[0148] In some embodiments, referring to FIG. 3 , the object detection model includes a first object detection model, which is trained based on battery cell sample images labeled with tab regions and tab misalignment regions.

[0149] It should be noted that the first target detection model is used to detect the position of the tab in the battery cell image and to identify the category of each target in the battery cell image.

[0150] In a specific implementation, the first target detection model is trained by using sample images of battery cells with tab regions and tab misalignment regions marked, so that the tab regions and tab misalignment regions can be detected by the first target detection model.

[0151] In some embodiments, the step of performing tab defect detection on battery cell images collected from multiple angles using an object detection model to obtain tab detection results includes:

[0152] Step S201 ′: performing image enhancement on the battery cell images collected from multiple angles through the input layer in the first object detection model to obtain battery cell enhanced images.

[0153] It should be noted that the input layer in the first target detection model is used for image enhancement, and the input battery cell images collected from multiple angles can be enhanced, specifically by random transformation, cropping, splicing, etc., so as to construct the battery cell images into images with richer and more complex feature and target combinations, namely, battery cell enhanced images.

[0154] Step S202 ′: extracting features from the battery cell enhanced image through the feature extraction layer in the first object detection model to obtain first tab extraction features.

[0155] It should be understood that after obtaining the battery cell enhanced image, the battery cell enhanced image is passed to the feature extraction layer in the first target detection model, and the convolution operation therein is used to extract the features of the battery cell enhanced image, thereby obtaining the first tab extraction feature.

[0156] Exemplarily, the first tab extraction feature is a feature that characterizes different targets in the battery cell image, such as the position of the tab, the position of the pole piece, etc.

[0157] Step S203 ′: performing feature fusion on the first tab extraction features through the feature fusion layer in the first target detection model to obtain the first tab fusion features.

[0158] In a specific implementation, after obtaining the first tab feature, the extracted first tab extraction features can be combined using the feature fusion layer in the first target detection model. The feature fusion layer uses upsampling, feature concatenation (concat) and other operations to transform the first tab extraction features. Finally, the features are divided into 3 outputs, each output representing the features of targets of different sizes summarized therein, thereby obtaining the first tab fusion feature.

[0159] It should be understood that targets of different sizes can be predicted by the first tab fusion feature, such as predicting the position of the tab and the pole piece.

[0160] Step S204 ′: performing tab location on the first tab fusion feature through the output layer in the first object detection model to obtain a tab detection area.

[0161] It can be understood that the first tab fusion feature can be transmitted to the output layer of the first target detection model for prediction, and the predicted target position and category can be output.

[0162] For example, the output layer in the first target detection model can be used to locate the tab by using the first tab fusion feature, thereby marking the tab detection area in the battery cell image and labeling the category as "tab".

[0163] Step S205 ′: obtaining a tab detection result through the tab detection area.

[0164] It should be noted that the tab detection results can be obtained through the tab detection area.

[0165] For example, the tab detection area is inspected to detect whether the tab has defects.

[0166] In the technical solution of the embodiment of the present application, the tab area in the battery cell image is detected by the first object detection model, the tab position can be determined first, and the tab can be roughly positioned, thereby improving the accuracy of tab defect detection.

[0167] In some embodiments, the tab detection area can be used to detect whether the tab is misaligned. Since the first target detection model is trained using battery cell sample images marked with tab misalignment areas, the first target detection model can also detect whether the tab is misaligned in the battery cell image.

[0168] As an example, the tab detection results obtained through the tab detection area include:

[0169] Check whether there is tab misalignment in the tab detection area;

[0170] When there is tab misalignment in the tab detection area, the tab detection result is determined to be tab misalignment.

[0171] It should be noted that whether there is tab misalignment can be marked during tab detection, so whether there is tab misalignment in the tab detection area can be detected.

[0172] As an example, the image of the tab detection area may be directly compared with an image without tab misalignment to determine whether tab misalignment exists in the tab detection area.

[0173] In a specific implementation, if there is no tab misalignment in the tab detection area, it is preliminarily determined that the tab has no defects, and the tab can be further inspected for defects.

[0174] In this embodiment, if tab misalignment is present in the tab inspection area, the tab inspection result can be set as tab misalignment, and the battery cell image can be labeled as such. The tab is then further inspected for other defects. This allows subsequent tab defect inspection to determine whether the tab defect is a surface defect or a defect in the tab misalignment area.

[0175] In this embodiment, the tab is detected to determine whether it is misaligned in the tab detection area obtained by the first target detection model, thereby enabling detection of tab misalignment defects and improving the tab detection range.

[0176] For example, when the tab detection result is tab misalignment, in order to improve subsequent detection efficiency, the tab misalignment area can be removed from the image. When tab misalignment exists in the tab detection area, after determining that the tab detection result is tab misalignment, the method further includes:

[0177] The tab misalignment region in the tab image is cropped to obtain a cropped tab image.

[0178] It should be noted that after the tab is positioned, specific defects of the tab need to be detected. In order to improve the detection efficiency, the tab image representing the tab detection area can be cropped, that is, the tab misalignment area in the tab image can be cropped, and the tab misalignment area can be cropped from the tab image to obtain a cropped tab image.

[0179] As an example, after obtaining the cropped tab image, more specific tab defect detection can be performed through the cropped tab image.

[0180] In this embodiment, when there is tab misalignment in the tab image, the tab misalignment region in the tab image is cropped to obtain a cropped tab image, thereby improving the accuracy of subsequent tab defect detection.

[0181] In some embodiments, the first object detection model can quickly and accurately detect the location and category of the tab in the battery cell image. Therefore, the process of establishing the first object detection model includes:

[0182] Acquire a battery cell sample image, wherein the battery cell sample image is marked with a tab region and a tab misalignment region;

[0183] The initial object detection network is trained using battery cell sample images to obtain a first object detection model.

[0184] It should be noted that the battery cell sample images may be images of a variety of different battery cells, images of battery cells in different forms, or images of battery cells during different production processes, which are not limited in this embodiment.

[0185] It should be noted that the battery cell sample image can be acquired through multiple angles by an image acquisition device. The battery cell sample image includes the tab portion and may also include other portions of the battery cell, such as the electrode portion.

[0186] It should be understood that the battery cell sample image is annotated with the tab area and the tab misalignment area, which can be annotated using professional annotation tools or manually assisted to improve the accuracy of the annotation.

[0187] In a specific implementation, the initial target detection network is the yolov5 target detection network, and may also be other networks that can implement target detection, which is not limited in this embodiment.

[0188] During model training, the battery cell sample images are first input into the input layer of the initial object detection network. At the input layer, the battery cell sample images undergo image enhancement preprocessing, such as random transformation, cropping, and splicing. This constructs the battery cell sample images into battery cell sample images with richer and more complex feature and target combinations, which serve as the model training set. This improves the trained model's recognition rate and robustness for abnormal and complex images. The enhanced battery cell sample images are then passed to the feature extraction layer of the yolov5 network for feature extraction and output of image features. Backone is used as the convolutional network. After extracting the features, the feature fusion layer fuses the features and transmits the fused features to the output layer for prediction. The predicted tab area and tab misalignment area are output and compared with the test set to obtain the loss. When the loss value is less than a set threshold, the model training is completed, resulting in the first object detection model. The goal of training is to ensure that the location and category prediction information of the tab area obtained for the input image is close to the annotated true information, thereby achieving the purpose of tab area detection.

[0189] In this embodiment, a battery cell sample image is obtained, and the battery cell sample image is annotated with the tab area and the tab misalignment area, so that the initial target detection network is trained through the battery cell sample image of the tab area and the tab misalignment area, so that the position of the tab part and the misalignment area obtained for the input battery cell image are close to the annotated real information, thereby achieving the purpose of detecting the tab position and the tab misalignment area.

[0190] In some embodiments, after obtaining the tab detection region, an image representing a tab portion in the battery cell image may be determined. Therefore, after obtaining the tab detection region, the defect detection method further includes:

[0191] The battery cell image is cropped according to the tab detection area to obtain the tab image.

[0192] It should be noted that the battery cell image can be cropped based on the tab detection area detected by the first target detection model to obtain a tab image containing only the tab, thereby obtaining an image of the specific tab area. This facilitates subsequent tab defect detection using the tab image and improves detection accuracy.

[0193] Exemplarily, the tab detection area is marked with the coordinate information of the tab area, so that the tab position can be quickly located and the image can be cropped to obtain an image containing only the tab.

[0194] In this embodiment, after the tab detection area is identified, the battery cell image is cropped into the tab image, so that the tab defects can be accurately identified directly through the tab image.

[0195] In some embodiments, in order to improve the accuracy of tab defect detection, defect detection may be performed on the tab image obtained by the first target detection model, thereby improving the accuracy of tab defect detection.

[0196] In some embodiments, referring to FIG. 4 , the object detection model includes a second object detection model, which is trained using tab sample images labeled with tab defects.

[0197] It should be noted that the second target detection model is obtained by training tab sample images marked with tab defects, so that tab defects and tab defect categories can be detected by the second target detection model.

[0198] Therefore, after obtaining the tab image, the tab image can be input into the second target detection model to perform tab specific defect detection. Therefore, the step of obtaining the tab detection result through the tab detection area includes:

[0199] Step S215 ′: inputting the tab image into the second object detection model.

[0200] It should be noted that the input layer in the target detection model will scale the image to speed up training and prediction. Therefore, a large amount of information will be lost in the image during the scaling process. In order to improve the detection accuracy, the tab image containing only the tab can be input into the second target detection model for defect detection.

[0201] It should be noted that the second target detection model and the first target detection model are both obtained by training the initial target detection network. The architecture of the model is the same, but the samples during training are different.

[0202] Step S225 ′: performing image enhancement on the tab image through the input layer in the second object detection model to obtain a tab enhanced image.

[0203] In a specific implementation, the lug image is first enhanced through the input layer in the second target detection model, for example, the lug image is cropped, spliced, and other operations are performed to obtain a lug enhanced image.

[0204] Step S235 ′: extracting features from the tab enhanced image through the feature extraction layer in the second object detection model to obtain a second tab extraction feature.

[0205] It should be understood that after obtaining the tab enhanced image, the tab enhanced image is passed to the feature extraction layer in the second target detection model, and the convolution operation therein is used to extract the features of the tab enhanced image, thereby obtaining the second tab extraction feature.

[0206] Exemplarily, the second tab extraction feature is a feature that characterizes different defects in the tab image, such as tab wrinkles, tab cracks, etc.

[0207] Step S245 ′: performing feature fusion on the second tab extraction features through the feature fusion layer in the second target detection model to obtain the second tab fusion features.

[0208] In a specific implementation, after obtaining the second tab feature, the extracted second tab extraction features can be combined using the feature fusion layer in the second target detection model. The feature fusion layer uses upsampling, feature concatenation (concat) and other operations to transform the second tab extraction features. Finally, the features are divided into 3 outputs, each output representing the features of targets of different sizes summarized therein, thereby obtaining the second tab fusion features.

[0209] It should be understood that targets of different sizes can be predicted through the second tab fusion feature, for example, specific tab defect categories and tab defect locations can be predicted.

[0210] Step S255 ′: performing tab defect detection on the second tab fusion feature through the output layer in the second object detection model to obtain a tab defect category.

[0211] It can be understood that the second tab fusion feature can be transmitted to the output layer of the second target detection model for prediction, and the predicted tab defect position and the specific defect category can be output.

[0212] For example, the output layer in the second target detection model can be used to locate the tab defect using the second tab fusion feature, thereby marking the tab defect position in the tab image and labeling the defect category, such as "tab crack" or "tab wrinkle".

[0213] Step S265 ′: obtaining tab detection results according to the tab defect category.

[0214] In a specific implementation, the tab defect category and the specific tab defect location can be used as the tab detection result.

[0215] In specific implementation, although defects are identified in the tab, if the defects are within the set standard value range, the final tab inspection result is that the tab does not have defects. Therefore, the tab can be further inspected according to the tab defect category to determine whether the tab has defects.

[0216] The tab inspection result may be that the tab has defects or that the tab does not have defects.

[0217] As shown in Figure 5, Figure 5 is a schematic diagram of the overall process of defect detection in one embodiment. By acquiring battery cell images acquired by the image acquisition module from multiple angles, the tab detection area in the battery cell image is located through the first target detection model, and it is identified whether there is tab misalignment. The tab detection area is input into the second target detection model for defect positioning and category recognition, thereby obtaining the tab defect position and tab defect category, and calculating the tab defect to determine whether it meets the standard, thereby determining whether the battery cell meets NG adjustment, and sending the result to the host computer.

[0218] In the technical solution of the embodiment of the present application, defects in the tab image are detected through a second target detection model, which can quickly detect the tab defect category, solve the tab misalignment problem that occurs in the battery cell production process, and improve the recognition rate of small defects and minor defects.

[0219] In some embodiments, after a tab defect is identified, it is possible to further detect whether the tab defect meets production requirements, that is, whether the tab defect is within an allowable error range. For example, further analysis can be performed by obtaining information such as the length or area of ​​the identified defective tab. Therefore, the step of obtaining a tab detection result based on the tab defect category includes:

[0220] Get the defect length corresponding to the tab defect category;

[0221] comparing the defect length to a preset length threshold;

[0222] When the defect length is greater than a preset length threshold, the tab inspection result is confirmed as a defect in the tab.

[0223] It should be noted that when detecting tab defects, the second target detection model will also mark the defect length corresponding to the tab defect. Therefore, the defect length corresponding to different tab defect categories can be obtained, thereby further confirming whether the tab has defects.

[0224] The preset length threshold is the allowable error threshold specified by the industry. If the defect length exceeds the preset length threshold, the tab is deemed to be defective. If the defect length does not exceed the preset length threshold, the tab is deemed not to be defective.

[0225] Therefore, the defect length can be compared with a preset length threshold. When the defect length is greater than the preset length threshold, the tab detection result is confirmed to be that the tab has a defect.

[0226] In this embodiment, the defect length corresponding to the tab defect category is obtained and the defect length is compared with a standard length threshold, thereby quickly determining whether the tab length has a defect.

[0227] In some embodiments, when the defect length is less than or equal to a preset length threshold, the tab inspection result is confirmed as no defect in the tab.

[0228] It should be noted that if the defect length is less than or equal to the preset length threshold, it is considered that the current tab defect is within the allowable error range, and the tab detection result is that there is no defect in the tab.

[0229] In this embodiment, by comparing the defect length with a standard length threshold, when the defect length is less than or equal to the standard length threshold, it is quickly determined that there is no defect in the tab length.

[0230] In some embodiments, further analysis can be performed by identifying the defective tab area to confirm whether the tab is defective. The step of obtaining the tab detection result based on the tab defect category includes:

[0231] Get the defect area corresponding to the tab defect category;

[0232] comparing the defect area with a preset area threshold;

[0233] When the defect area is greater than a preset area threshold, the tab inspection result is confirmed as a defect in the tab.

[0234] It should be noted that when detecting tab defects, the second target detection model will also mark the defect area corresponding to the tab defect. Therefore, the defect area corresponding to different tab defect categories can be obtained, thereby further confirming whether the tab has defects.

[0235] The preset area threshold is the allowable error threshold specified by the industry. If the defect area exceeds the preset area threshold, the tab is deemed to be defective. If the defect length does not exceed the preset area threshold, the tab is deemed not to be defective.

[0236] Therefore, the defect area can be compared with a preset area threshold. When the defect area is greater than the preset area threshold, the tab inspection result is confirmed to be that the tab has a defect.

[0237] In this embodiment, the defect area corresponding to the tab defect category is obtained and compared with the standard area threshold, thereby quickly determining whether there is a defect on the tab area.

[0238] In some embodiments, when the defect area is less than or equal to a preset area threshold, the tab inspection result is confirmed as no defect in the tab.

[0239] It should be noted that if the defect area is less than or equal to the preset area threshold, it is considered that the current tab defect is within the allowable error range, and the tab detection result is that there is no defect in the tab.

[0240] In the technical solution of the embodiment of the present application, by comparing the defect area with the standard area threshold, when the defect area is less than or equal to the standard area threshold, it is quickly determined that there is no defect in the tab area.

[0241] In some embodiments, the second object detection model can quickly and accurately detect the location and type of tab defects in the tab image. Therefore, the process of establishing the second object detection model includes:

[0242] Acquire a tab sample image, wherein the tab sample image is marked with a tab defect;

[0243] The initial object detection network is trained using the ear sample images to obtain the second object detection model.

[0244] It should be noted that the tab sample images may be tab images collected from a variety of different battery cells, tab images of battery cells in different forms, or tab images of battery cells during different production processes, which is not limited in this embodiment.

[0245] It should be noted that the tab sample image can be obtained by performing tab area detection on the battery cell sample image using the first target detection model, or can be obtained by directly capturing the tab portion using an image acquisition device, which is not limited in this embodiment.

[0246] It should be understood that the tab sample image is annotated with tab defects, including the tab defect location and tab defect category, which can be annotated using professional annotation tools or manually assisted to improve the accuracy of the annotation.

[0247] In a specific implementation, the initial target detection network is the yolov5 target detection network, and may also be other networks that can implement target detection, which is not limited in this embodiment.

[0248] During model training, the tab sample images are first input into the input layer of the initial object detection network. Image enhancement preprocessing, such as random transformation, cropping, and splicing, is performed on the tab sample images at the input layer. This constructs the tab sample images into more complex and richer feature and target combinations, serving as the model training set. This improves the trained model's recognition rate and robustness for abnormal and complex images. The enhanced tab sample images are then passed to the feature extraction layer of the yolov5 network for feature extraction and output of image features. Backone is used as the convolutional network. After extracting the features, the feature fusion layer is used for feature fusion, and the fused features are fed to the output layer for prediction. The predicted tab defect locations and categories are output and compared with the test set to obtain a loss. When the loss value is less than a set threshold, model training is completed, resulting in the second object detection model. The goal of training is to ensure that the predicted tab defect locations and categories for the input images are close to the annotated true information, thereby achieving the purpose of tab defect detection.

[0249] In the technical solution of the embodiment of the present application, by obtaining a sample image of the tab, the sample image of the tab is marked with tab defects, and the initial target detection network is trained using the sample image of the tab. Since a large amount of information will be lost during the scaling process of the image, the second target detection model is obtained by training using the sample image of the tab to improve the model training accuracy.

[0250] In some embodiments, if a tab defect is identified, specific information needs to be provided as a prompt or warning. Therefore, the steps of completing defect detection based on the tab inspection results include:

[0251] When the tab inspection result shows that the tab has defects, the tab defect category and tab defect location are output;

[0252] Determine that the battery cell corresponding to the tab does not meet the production requirements and remove the battery cell.

[0253] It should be noted that if the tab inspection result shows that the tab is defective, the tab defect category and tab defect location can be output to the host computer.

[0254] It is also marked that the battery cell corresponding to the defective tab does not meet the production requirements, and the battery cell can be taken out, for example, the battery cell can be directly discharged into the NG pool.

[0255] The overall inspection process can be controlled within the cycle time of the battery cell, enabling real-time detection of tab defects without affecting battery cell production.

[0256] In this embodiment, when the tab detection result shows that the tab is defective, the specific tab defect category and tab defect location can be output, so that the battery cell to which the defective tab belongs is judged as not meeting production requirements and is removed, thereby realizing intelligent control of the production line and effectively ensuring the production line rhythm.

[0257] To achieve the above-mentioned purpose, an embodiment of the present application further proposes a defect detection system 1, as shown in FIG6 , which is a schematic structural diagram of the defect detection system 1 in an embodiment of the present application.

[0258] In some embodiments, the defect detection system 1 includes:

[0259] Image acquisition module 10, used for acquiring images of battery cells including the tabs at multiple angles;

[0260] The host computer 20 is used to confirm whether the tab has defects based on the battery cell images collected from multiple angles.

[0261] It should be noted that the image acquisition module 10 may be a device with an image acquisition function, such as a camera, an image sensor, etc., or may be other devices capable of image acquisition, which is not limited in this embodiment.

[0262] The number of image acquisition modules 10 may include one or more. For example, the number of image acquisition modules is multiple and they can be arranged around the battery cell, so as to capture images of the battery cell including the tab area from multiple angles.

[0263] It should be noted that the image acquisition module 10 can be communicatively connected to the host computer 20 , and the image acquisition module 10 sends the battery cell images including the tab portion acquired from multiple angles to the host computer 20 .

[0264] In a specific implementation, the host computer 20 may receive a battery cell image and process the battery cell image to confirm whether the tab has defects.

[0265] In the technical solution of the embodiment of the present application, the image acquisition module is used to capture images of the battery cell including the tab area at multiple angles, so that a complete battery cell image can be obtained, the accuracy of detection can be improved, and the host computer can be used to confirm whether the tab has defects, thereby improving the efficiency and accuracy of tab defect detection and the recognition rate of minor tab defects.

[0266] In some embodiments, as shown in FIG7 , FIG7 is another structural diagram of a defect detection system in an embodiment of the present application.

[0267] In some embodiments, the image acquisition module 10 includes two, each image acquisition module 10 includes at least two image acquisition devices 101, and the two image acquisition modules 10 are respectively located at the end of the anode ear 21 and the end of the cathode ear 22;

[0268] The image acquisition module is used to perform multi-angle image acquisition on the battery cell including the tab area, obtain the multi-angle acquired battery cell images and send them to the host computer.

[0269] It should be noted that the tab 2 on the battery cell includes an anode tab 21 and a cathode tab 22. Therefore, an image acquisition module 10 for acquiring an image of the anode tab 21 and an image acquisition module 10 for acquiring an image of the cathode tab 22 can be respectively provided, thereby synchronously acquiring images of the anode tab 21 and the cathode tab 22.

[0270] In a specific implementation, the image acquisition module 10 includes two, which are respectively arranged at the end of the anode ear 21 and the end of the cathode ear 22, and each image acquisition device 10 includes at least two image acquisition devices 101. The two image acquisition devices 101 can be set at different angles. As shown in Figure 7, one image acquisition device 101 is set at the top and the other image acquisition device 101 is set at the bottom, so as to capture images of the anode ear 21 or the cathode ear 22 in the battery cell from different angles.

[0271] Exemplarily, the image acquisition device 101 may be a device with an image acquisition function, such as a camera or a scanner.

[0272] In some embodiments, the anode ears include two, and the cathode ears include two; the image acquisition module 10 includes a first image acquisition module and a second image acquisition module (not shown in the figure);

[0273] The two image acquisition devices 101 in the first image acquisition module 10 are respectively located at the ends of the two anode ears, and the two image acquisition devices 101 in the second image acquisition module 10 are respectively located at the ends of the two cathode ears.

[0274] In a specific implementation, two anode tabs and two cathode tabs are provided on the battery cell, and each tab is provided with two image acquisition devices 101 for image acquisition, thereby obtaining battery cell images acquired from multiple angles.

[0275] The anode ear or cathode ear is captured at multiple angles by multiple image acquisition devices, thereby completely capturing the surface of the pole ear, obtaining a complete pole ear image, and improving the accuracy of detection.

[0276] Optionally, in order to improve the clarity of the captured image, fill light may be provided during image capture. Therefore, the defect detection system further includes at least one light source 30, which is disposed on both sides of the battery cell opposite to the image capture module 10.

[0277] The at least one light source 30 is used to provide light when the image acquisition module 10 acquires an image of the battery cell including the tab portion.

[0278] It should be noted that when the image acquisition module 10 acquires the battery cell image, fill light can be provided by a light source, thereby improving the shooting effect.

[0279] By further arranging a light source in the system, fill light is provided when the image acquisition device acquires the battery cell image, so that a clear battery cell image including the tab portion can be captured, thereby improving the shooting effect.

[0280] In the technical solution of the embodiment of the present application, two image acquisition modules are used to respectively acquire images of battery cells including anode ears and images of battery cells including cathode ears, so as to perform ear defect detection separately, thereby avoiding the mixing of anode ear images and cathode ear images to affect the detection results.

[0281] The above are only preferred embodiments of the present application and do not limit the patent scope of the present application. All equivalent structural transformations made based on the contents of the present application specification and drawings, or direct / indirect applications in other related technical fields, are included in the patent protection scope of the present application.

Claims

1. A defect detection method, wherein: The defect detection method comprises: Acquire battery cell images collected from multiple angles, wherein the battery cell images include the tabs of the battery cells; Performing tab defect detection on battery cell images collected from multiple angles through a target detection model to obtain tab detection results, wherein the target detection model is used for tab positioning and tab defect identification; Defect detection is completed through the tab detection result.

2. The defect detection method according to claim 1, wherein: The target detection model includes an input layer, a feature extraction layer, a feature fusion layer and an output layer; The method of performing tab defect detection on battery cell images collected from multiple angles through the target detection model to obtain tab detection results includes: The battery cell images collected from multiple angles are enhanced through the input layer in the target detection model to obtain the battery cell enhanced image; Extracting features from the battery cell enhanced image through a feature extraction layer in the target detection model to obtain extracted features; Performing feature fusion on the extracted features through a feature fusion layer in the target detection model to obtain fused features; The fused features are used to perform lug position detection or lug defect detection through the output layer in the target detection model to obtain a lug detection result.

3. The defect detection method according to claim 2, wherein: The step of extracting features from the battery cell enhanced image through the feature extraction layer in the target detection model to obtain extracted features includes: Multi-scale feature extraction is performed on the battery cell enhanced image through a convolutional neural network in a feature extraction layer in the target detection model to obtain multi-scale extracted features.

4. The defect detection method according to claim 2, wherein: The target detection model includes a first target detection model, which is obtained by training a battery cell sample image with a tab region and a tab misalignment region marked; The method of performing tab defect detection on battery cell images collected from multiple angles through the target detection model to obtain tab detection results includes: Performing image enhancement on the battery cell images collected from multiple angles through the input layer in the first target detection model to obtain a battery cell enhanced image; Performing feature extraction on the battery cell enhanced image through the feature extraction layer in the first target detection model to obtain a first tab extraction feature; Performing feature fusion on the first tab extraction features through the feature fusion layer in the first target detection model to obtain a first tab fusion feature; Performing lug positioning on the first lug fusion feature through the output layer in the first target detection model to obtain a lug detection area; The tab detection result is obtained through the tab detection area.

5. The defect detection method according to claim 4, wherein: The performing lug positioning on the first lug fusion feature through the output layer in the first target detection model to obtain a lug detection area includes: The first tab fusion feature is used to locate the tab through the output layer in the first target detection model, and the tab detection area in the battery cell image is marked and labeled with the corresponding category to obtain the tab detection area.

6. The defect detection method according to claim 4, wherein: The defect detection method further includes: The battery cell image is cropped according to the tab detection area to obtain a tab image.

7. The defect detection method according to claim 6, wherein: The target detection model further includes a second target detection model, which is obtained by training a sample image of a tab annotated with a tab defect; The step of obtaining the tab detection result through the tab detection area includes: Inputting the tab image into the second object detection model; Performing image enhancement on the tab image through the input layer in the second object detection model to obtain a tab enhanced image; Performing feature extraction on the tab enhanced image through the feature extraction layer in the second object detection model to obtain a second tab extraction feature; Performing feature fusion on the second tab extraction features through the feature fusion layer in the second object detection model to obtain second tab fusion features; Performing tab defect detection on the second tab fusion feature through the output layer in the second object detection model to obtain a tab defect category; The tab detection result is obtained according to the tab defect category.

8. The defect detection method according to claim 7, wherein: The step of obtaining the tab detection result according to the tab defect category includes: Obtaining the defect length corresponding to the tab defect category; comparing the defect length with a preset length threshold; When the defect length is greater than the preset length threshold, the tab detection result is confirmed as a defect in the tab.

9. The defect detection method according to claim 8, wherein: The defect detection method further includes: When the defect length is less than or equal to the preset length threshold, it is confirmed that the tab detection result is that the tab has no defects.

10. The defect detection method according to claim 7, wherein: The step of obtaining the tab detection result according to the tab defect category includes: Obtaining a defect area corresponding to the tab defect category; comparing the defect area with a preset area threshold; When the defect area is greater than the preset area threshold, the tab detection result is confirmed as a defect in the tab.

11. The defect detection method according to claim 10, wherein: The defect detection method further includes: When the defect area is less than or equal to the preset area threshold, the tab inspection result is confirmed to be that the tab has no defects.

12. The defect detection method according to claim 2, wherein: The step of performing lug position detection or lug defect detection on the fused features through the output layer in the target detection model to obtain a lug detection result includes: Performing lug position detection or lug defect detection on the fused features through the output layer in the target detection model to obtain the lug position and the predicted target category; The tab position and the predicted target category are used as tab detection results.

13. The defect detection method according to claim 6, wherein: The step of obtaining the tab detection result through the tab detection area includes: Detecting whether there is a tab misalignment in the tab detection area; When there is a tab misalignment in the tab detection area, the tab detection result is determined to be a tab misalignment.

14. The defect detection method according to claim 13, wherein: The detecting whether there is a tab misalignment in the tab detection area includes: The image of the tab detection area is compared with an image without tab misalignment, and whether there is tab misalignment in the tab detection area is determined according to the comparison result.

15. The defect detection method according to claim 13, wherein: The defect detection method further includes: The tab misalignment region in the tab image is cropped to obtain a cropped tab image.

16. The defect detection method according to any one of claims 1 to 15, wherein: The target detection model includes a first target detection model and a second target detection model. The target detection model is used to perform tab defect detection on battery monomer images collected from multiple angles to obtain tab detection results, including: Locating the tab detection area in the battery cell image by using the first target detection model, and identifying whether there is tab misalignment, to obtain the tab detection area; Inputting the cropped tab image in the tab detection area into the second object detection model for defect location and category recognition to obtain the tab defect position and tab defect category; The tab detection result is obtained according to the tab defect position and the tab defect category.

17. The defect detection method according to any one of claims 1 to 16, wherein: Before the target detection model is used to detect the tab defects of the battery cell images collected from multiple angles and the tab detection results are obtained, the method further includes: Acquire a battery cell sample image, wherein the battery cell sample image is marked with a tab region and a tab misalignment region; The initial target detection network is trained using the battery cell sample images to obtain a first target detection model.

18. The defect detection method according to any one of claims 1 to 17, wherein: Before the target detection model is used to detect the tab defects of the battery cell images collected from multiple angles and the tab detection results are obtained, the method further includes: Acquire a tab sample image, wherein the tab sample image is marked with a tab defect; The initial target detection network is trained using the tab sample images to obtain a second target detection model.

19. The defect detection method according to any one of claims 1 to 18, wherein: The defect detection is completed by using the tab detection result, including: When the tab detection result shows that the tab has a defect, outputting the tab defect category and the tab defect position; It is determined that the battery cell corresponding to the tab does not meet the production requirements, and the battery cell is discarded.

20. A defect detection system, wherein: The defect detection system comprises: An image acquisition module, used to acquire images of battery cells including the lugs at multiple angles; The host computer is used to confirm whether there are defects in the tab based on the battery cell images collected from multiple angles.

21. The defect detection system of claim 20, wherein: The image acquisition modules include two, each of which includes at least two image acquisition devices, and the two image acquisition modules are respectively located at the end of the anode ear and the end of the cathode ear; The image acquisition module is used to perform multi-angle image acquisition on the battery cell image including the tab portion, obtain the multi-angle acquired battery cell image and send it to the host computer.

22. The defect detection system of claim 21, wherein: The anode ears include two, and the cathode ears include two; the image acquisition module includes a first image acquisition module and a second image acquisition module; The two image acquisition devices in the first image acquisition module are respectively located at the ends of the two anode ears, and the two image acquisition devices in the second image acquisition module are respectively located at the ends of the two cathode ears.

23. The defect detection system according to any one of claims 20 to 22, wherein: The system further comprises at least one light source, wherein at least one light source and the image acquisition module are arranged on two sides of the battery cell opposite to each other; At least one of the light sources is used to provide light when the image acquisition module acquires an image of a battery cell including a tab portion.

Images