Battery anomaly detection system, detection method, storage medium, and battery swap station

Abstract

The application relates to the technical field of power batteries, and in particular to a battery abnormality detection system, a battery abnormality detection method, a computer device, a computer storage medium and a battery swap station. The battery abnormality detection system according to an aspect of the application comprises: a communication module configured to acquire a battery image collected by an image collection device, battery identification information and identification information of the image collection device; an abnormality detection module configured to acquire a battery image feature data set from a cloud database according to the battery identification information and the identification information of the image collection device received from the communication module, and generate a battery abnormality detection result based on at least the collected battery image and the battery image feature data set; and a judgment module configured to judge whether an abnormal area is a repeatedly detected abnormal area in response to the battery abnormality detection result indicating that the battery has the abnormal area, and send the battery abnormality detection result in response to judging that the abnormal area is not the repeatedly detected abnormal area.

Description

Technical Field

[0001] This application relates to the field of battery technology, and more specifically to battery anomaly detection systems, battery anomaly detection methods, computer equipment for implementing the methods, computer storage media, and battery swapping stations. Background Technology

[0002] In electric vehicles, the power battery is a crucial component, and its performance affects the vehicle's driving range and safety. During driving, uneven roads, unusual protrusions, or collisions can cause damage to the battery surface on the vehicle's underside, or allow foreign objects to adhere to it, thus impacting battery safety.

[0003] Currently, image edge detection methods based on logical rules or object detection methods based on deep learning are generally used to detect battery damage or defects. While image edge detection methods based on logical rules offer good interpretability and flexible configuration of logical parameters, their adaptability and robustness are poor, making them unsuitable for battery images acquired in open environments. Because the battery surface is exposed at the bottom of the electric vehicle and is often contaminated with water stains, dirt, snow, oil, and other foreign objects, image edge detection methods based on logical rules struggle to accurately distinguish between damage and foreign objects, leading to false detections. Object detection methods based on deep learning require a large number of available anomalous samples for training, but most defect scenarios lack a rich pool of such samples. This makes the algorithm prone to overfitting during training, resulting in low accuracy in identifying obvious but uncommon anomalous regions, and the overall algorithm's self-iterative ability is also poor.

[0004] Furthermore, because the battery surface has a protective metal layer, minor damage does not require repair. After monitoring and alarm activation and manual confirmation, the slightly damaged battery is deemed risk-free. However, damage in the same location may be detected repeatedly and trigger repeated alarms. These repeated alarms cause on-site maintenance personnel to perform repetitive tasks, significantly impacting work efficiency. Summary of the Invention

[0005] To address or at least alleviate one or more of the above problems, the following technical solutions are provided.

[0006] According to a first aspect of this application, a battery anomaly detection system is provided, the system comprising: a communication module configured to acquire a battery image, battery identification information, and identification information of the image acquisition device acquired by an image acquisition device; an anomaly detection module configured to acquire a battery image feature dataset from a cloud database based on the battery identification information and the identification information of the image acquisition device received from the communication module, and to generate a battery anomaly detection result based at least on the acquired battery image and the battery image feature dataset; and a judgment module configured to determine whether the anomaly region is a repeatedly detected anomaly region in response to the battery anomaly detection result indicating that an anomaly region exists in the battery, and to send the battery anomaly detection result in response to the determination that the anomaly region is not a repeatedly detected anomaly region.

[0007] According to an embodiment of the battery anomaly detection system of this application, the battery image feature dataset includes: first battery image feature data generated based on semantic vectors of multiple regions of the battery image stored in the identification information of the image acquisition device; second battery image feature data generated based on semantic vectors of multiple regions of the battery image stored in the battery identification information; third battery image feature data generated from semantic vectors of threatening anomaly regions; and fourth battery image feature data generated from semantic vectors of non-threatening anomaly regions.

[0008] According to one embodiment or any of the above embodiments of the battery anomaly detection system, the anomaly detection module is further configured to: send one or more of the battery anomaly detection results, the acquired battery image, the battery identification information, and the identification information of the image acquisition device to the cloud database for storage.

[0009] According to one embodiment or any of the above embodiments of the battery anomaly detection system, the cloud database stores one or more of the following data: the battery image feature dataset, original image data, image size data, pixel position data corresponding to the threat anomaly region, threat anomaly region and its label, non-threat anomaly region and its label, battery anomaly detection result, the acquired battery image, battery identification information, identification information of the image acquisition device, and historical battery anomaly detection results.

[0010] According to one embodiment or any of the above embodiments of the present application, the battery anomaly detection system includes: an anomaly warning unit configured to generate an anomaly warning region based at least on the acquired battery image, the first battery image feature data, and the second battery image feature data; and an anomaly diagnosis unit configured to generate a threat anomaly score for the anomaly warning region based at least on the anomaly warning region, the third battery image feature data, and the fourth battery image feature data.

[0011] According to an embodiment of this application or any of the above embodiments, the battery anomaly detection system includes: a target detection subunit configured to detect the acquired battery image and generate a first anomaly warning region in response to detecting an abnormal target in the acquired battery image that is greater than an anomaly warning score threshold; a first comparison subunit configured to process the acquired battery image to obtain semantic vectors of multiple regions of the acquired battery image and compare the semantic vectors of the multiple regions of the acquired battery image with the first battery image feature data to generate a second anomaly warning region; and a second comparison subunit configured to process the acquired battery image to obtain semantic vectors of multiple regions of the acquired battery image and compare the semantic vectors of the multiple regions of the acquired battery image with the second battery image feature data to generate a third anomaly warning region.

[0012] According to one embodiment or any of the above embodiments of the battery anomaly detection system, the anomaly warning unit is further configured to generate the anomaly warning region based on the first anomaly warning region, the second anomaly warning region and the third anomaly warning region.

[0013] According to an embodiment of this application or any of the above embodiments, the battery anomaly detection system includes: a classification subunit configured to process the anomaly warning region to generate an anomaly score for the anomaly warning region; a first comparison subunit configured to process the anomaly warning region to obtain a semantic vector of the anomaly warning region and compare the semantic vector of the anomaly warning region with the third battery image feature data to generate a first similarity score for the anomaly warning region; and a second comparison subunit configured to process the anomaly warning region to obtain a semantic vector of the anomaly warning region and compare the semantic vector of the anomaly warning region with the fourth battery image feature data to generate a second similarity score for the anomaly warning region.

[0014] According to one embodiment or any of the above embodiments of the present application, the battery anomaly detection system, wherein the anomaly diagnosis unit is further configured to: perform weighted processing on the anomaly score, the first similarity score and the second similarity score to generate a threat anomaly score for the anomaly warning region; and, in response to the threat anomaly score for the anomaly warning region being greater than a threat anomaly score threshold, use the anomaly warning region as the anomaly region indicating the presence of the battery in the battery anomaly detection result.

[0015] According to one embodiment or any of the above embodiments of the battery anomaly detection system, the battery anomaly detection result includes one or more of the following: whether the battery has the abnormal region, the number of the abnormal regions, and the threat anomaly score corresponding to the abnormal region.

[0016] According to one embodiment or any of the above embodiments of the present application, the battery anomaly detection system further includes: an alarm module configured to generate display data based on the battery anomaly detection result, the display data being used to display the anomaly area on the acquired battery image.

[0017] According to one embodiment or any of the above embodiments of the present application, the battery anomaly detection system further includes: an anomaly detection training module configured to update the anomaly detection module based on one or more of the following data in the cloud database: original image data, image size data, pixel position data corresponding to threat anomaly regions, threat anomaly regions and their labels, non-threat anomaly regions and their labels.

[0018] According to one embodiment or any of the above embodiments of the battery anomaly detection system of this application, the determination module is further configured to determine whether the anomaly region is a repeatedly detected anomaly region based on one or more of the following comparisons: a comparison between the number of anomaly regions and the number of anomaly regions detected in the previous detection; a comparison between the minimum repetition probability of the anomaly region and a probability threshold; and a comparison between the similarity between the semantic vector of the anomaly region and the semantic vector of the previously detected anomaly region and a similarity threshold.

[0019] According to one embodiment of the present application or any of the above embodiments, the battery anomaly detection system wherein the image acquisition device is located at a battery swapping station and configured to acquire battery images during battery swapping operations.

[0020] According to a second aspect of this application, a battery anomaly detection method is provided, the method comprising: acquiring a battery image, battery identification information, and identification information of the image acquisition device acquired by an image acquisition device; acquiring a battery image feature dataset from a cloud database based on the battery identification information and the identification information of the image acquisition device, and generating a battery anomaly detection result based at least on the acquired battery image and the battery image feature dataset; determining whether the anomaly region is a repeatedly detected anomaly region in response to the battery anomaly detection result indicating the presence of an anomaly region in the battery, and sending the battery anomaly detection result in response to determining that the anomaly region is not a repeatedly detected anomaly region.

[0021] According to an embodiment of the battery anomaly detection method of this application, the battery image feature dataset includes: first battery image feature data generated based on semantic vectors of multiple regions of the battery image stored based on the identification information of the image acquisition device; second battery image feature data generated based on semantic vectors of multiple regions of the battery image stored based on the battery identification information; third battery image feature data generated from semantic vectors of threatening anomaly regions; and fourth battery image feature data generated from semantic vectors of non-threatening anomaly regions.

[0022] The battery anomaly detection method according to one embodiment or any of the above embodiments of this application further includes: sending one or more of the battery anomaly detection result, the acquired battery image, the battery identification information, and the identification information of the image acquisition device to the cloud database for storage.

[0023] According to one embodiment or any of the above embodiments of the battery anomaly detection method, the cloud database stores one or more of the following data: the battery image feature dataset, original image data, image size data, pixel position data corresponding to the threat anomaly region, threat anomaly region and its label, non-threat anomaly region and its label, the battery anomaly detection result, the acquired battery image, the battery identification information, the identification information of the image acquisition device, and historical battery anomaly detection results.

[0024] The battery anomaly detection method according to one embodiment or any of the above embodiments of this application further includes: generating display data based on the battery anomaly detection result, wherein the display data is used to display the anomaly area on the acquired battery image.

[0025] According to one embodiment or any of the above embodiments of the present application, the battery anomaly detection method further includes: updating the anomaly detection module based on one or more of the following data in the cloud database: original image data, image size data, pixel position data corresponding to the threat anomaly region, the threat anomaly region and its label, and the non-threat anomaly region and its label.

[0026] According to one embodiment or any of the above embodiments of the present application, the battery anomaly detection method, determining whether the anomaly region is a repeatedly detected anomaly region includes determining whether the anomaly region is a repeatedly detected anomaly region based on one or more of the following comparisons: a comparison between the number of anomaly regions and the number of anomaly regions detected in the previous detection; a comparison between the minimum repetition probability of the anomaly region and a probability threshold; and a comparison between the similarity between the semantic vector of the anomaly region and the semantic vector of the previously detected anomaly region and a similarity threshold.

[0027] According to a third aspect of this application, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the computer program to implement the steps of the battery anomaly detection method according to a second aspect of this application.

[0028] According to a fourth aspect of this application, a computer storage medium is provided, which includes instructions that, when executed, perform the steps of the battery anomaly detection method according to a second aspect of this application.

[0029] According to a fifth aspect of this application, a battery swapping station is provided, comprising: an image acquisition device configured to acquire battery images; a communication device configured to transmit the acquired battery images to a battery anomaly detection system according to a first aspect of this application and to receive display data from the battery anomaly detection system; and a display device configured to display a battery anomaly area on the acquired battery images based on the display data.

[0030] The battery anomaly detection scheme according to one or more embodiments of this application can accurately identify abnormal areas in the battery through the anomaly detection module, and can prevent repeated detection of the same abnormal area through the judgment module. Therefore, the accuracy of battery anomaly detection is improved, thereby ensuring battery safety and enhancing operational efficiency.

[0031] Appendix

[0032] Figure caption

[0033] The above and / or other aspects and advantages of this application will become clearer and more readily understood from the following description taken in conjunction with the accompanying drawings, in which the same or similar elements are denoted by the same reference numerals. The drawings include:

[0034] Figure 1 A schematic block diagram of a battery anomaly detection system according to one or more embodiments of this application is shown.

[0035] Figure 2 A schematic block diagram of a battery anomaly detection system according to one or more embodiments of this application is shown.

[0036] Figure 3 A schematic block diagram of a cloud database according to one or more embodiments of this application is shown.

[0037] Figure 4 A schematic diagram of semantic vectors for generating a battery image of multiple regions according to one or more embodiments of this application is shown.

[0038] Figure 5 A schematic diagram illustrating the generation of semantic vectors for threat-anomalous regions and non-threat-anomalous regions according to one or more embodiments of this application is shown.

[0039] Figure 6 A schematic diagram of an anomaly detection module according to one or more embodiments of this application is shown.

[0040] Figure 7 A schematic diagram of a comparison subunit of an anomaly warning unit according to one or more embodiments of this application is shown.

[0041] Figure 8 A schematic diagram of a comparison subunit of an anomaly diagnosis unit according to one or more embodiments of this application is shown.

[0042] Figure 9 A flowchart of a battery anomaly detection method according to one or more embodiments of this application is shown.

[0043] Figure 10 A flowchart illustrating a method for determining whether an abnormal region is a repeatedly detected abnormal region according to one or more embodiments of this application is shown.

[0044] Figure 11 A block diagram of a computer device according to one or more embodiments of this application is shown. Detailed Implementation

[0045] The present application will now be described more fully with reference to the accompanying drawings, which illustrate exemplary embodiments thereof. However, the present application may be implemented in various forms and should not be construed as being limited to the embodiments given herein. The foregoing embodiments are intended to make the disclosure herein complete and thorough, so as to more fully convey the scope of protection of the present application to those skilled in the art.

[0046] In this specification, terms such as “comprising” and “including” indicate that, in addition to having the units and steps that are directly and explicitly stated in the specification and claims, the technical solution of this application does not exclude the presence of other units and steps that are not directly or explicitly stated.

[0047] Unless otherwise specified, terms such as “first” and “second” do not indicate the order of units in terms of time, space, size, etc., but are merely used to distinguish between units.

[0048] In the following, various exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings.

[0049] Figure 1 A schematic block diagram of a battery anomaly detection system according to one or more embodiments of this application is shown.

[0050] like Figure 1 As shown, the battery anomaly detection system 100 includes a communication module 110, an anomaly detection module 120, and a judgment module 130.

[0051] The communication module 110 is configured to acquire battery images, battery identification information, and image acquisition device identification information acquired by the image acquisition device, and then send these acquired battery images, battery identification information, and image acquisition device identification information to the anomaly detection module 120. Optionally, the image acquisition device can be located at the battery swapping station and configured to acquire battery images during the battery swapping operation, for example, acquiring images of the battery surface at the start and end of the swapping process. The battery identification information can be understood as battery identity information used to distinguish different batteries, and the image acquisition device identification information can be understood as image acquisition device identity information used to distinguish different image acquisition devices.

[0052] The anomaly detection module 120 is configured to retrieve a battery image feature dataset from a cloud database based on battery identification information received from the communication module 110 and identification information of the image acquisition device, and to generate a battery anomaly detection result based at least on the acquired battery image and the retrieved battery image feature dataset. Optionally, the battery image feature dataset may include: first battery image feature data generated based on semantic vectors of multiple regions of the battery image stored based on the identification information of the image acquisition device; second battery image feature data generated based on semantic vectors of multiple regions of the battery image stored based on the identification information of the battery; third battery image feature data generated by semantic vectors of threat-anomalous regions; and fourth battery image feature data generated by semantic vectors of non-threat-anomalous regions. Threat-anomalous regions may include, but are not limited to, damaged areas, dented areas, foreign object areas, and other areas on the battery surface that pose a threat to the battery. Non-threat-anomalous regions may include, but are not limited to, water stains, oil stains, and other contaminant areas on the battery surface that do not pose a threat to the battery. Optionally, the battery anomaly detection result may include one or more of the following: whether the battery has the anomalous region, the number of anomalous regions, and a threat anomaly score corresponding to the anomalous region.

[0053] Optionally, the anomaly detection module 120 can also be configured to send one or more of the following to a cloud database for storage: battery anomaly detection results, acquired battery images, battery identification information, and identification information of the image acquisition device. Optionally, the anomaly detection module 120 can establish various forms of communication connections with the cloud database. Optionally, the data stored in the cloud database includes, but is not limited to, battery image feature datasets, raw image data, image size data, pixel position data corresponding to threat anomaly areas, threat anomaly areas and their labels, non-threat anomaly areas and their labels, battery anomaly detection results, acquired battery images, battery identification information, identification information of the image acquisition device, and historical battery anomaly detection results.

[0054] The determination module 130 is configured to determine whether an abnormal region is a repeatedly detected abnormal region in response to a battery anomaly detection result indicating the existence of an abnormal region in the battery, and to send a battery anomaly detection result in response to determining that the abnormal region is not a repeatedly detected abnormal region. Optionally, the determination module 130 may be configured to determine whether an abnormal region is a repeatedly detected abnormal region based on one or more of the following comparisons: a comparison between the number of abnormal regions and the number of abnormal regions detected in the previous detection; a comparison between the minimum repetition probability of the abnormal region and a probability threshold; and a comparison between the similarity between the semantic vector of the abnormal region and the semantic vector of the previously detected abnormal region and a similarity threshold.

[0055] The battery anomaly detection system according to one aspect of this application can accurately identify abnormal areas in the battery through an anomaly detection module, and can prevent repeated detection of the same abnormal area through a judgment module. This improves the accuracy of battery anomaly detection, ensuring battery safety and enhancing operational efficiency.

[0056] Figure 2 A schematic block diagram of a battery anomaly detection system according to one or more embodiments of this application is shown.

[0057] like Figure 2 As shown, the battery anomaly detection system 200 includes a communication module 210, an anomaly detection module 220, a judgment module 230, an alarm module 240, a cloud storage module 250, and an anomaly detection training module 260.

[0058] The communication module 210 is configured to acquire battery images, battery identification information, and identification information of the image acquisition device acquired by the image acquisition device, and send the acquired battery images, battery identification information, and image acquisition device identification information to the anomaly detection module 220. Optionally, the image acquisition device can be located at the battery swapping station and configured to acquire battery images during the battery swapping operation, for example, acquiring images of each surface of the battery at the start and end of the battery swapping. Optionally, the communication module 210 can also be configured to acquire the surface type (e.g., top surface, bottom surface, side surface, etc.) and battery swapping service identification information of the acquired battery images, and send the acquired surface type and battery swapping service identification information of the acquired battery images to the anomaly detection module 220.

[0059] The anomaly detection module 220 is configured to send the acquired battery image, battery identification information, image acquisition device identification information, and optional surface type and battery swapping service identification information of the acquired battery image received from the communication module 210 to the cloud storage module 250. Based on the battery identification information and image acquisition device identification information, it retrieves a battery image feature dataset from the cloud storage module 250, and generates a battery anomaly detection result based on one or more of the acquired battery image, the retrieved battery image feature dataset, the battery identification information, the image acquisition device identification information, and optional surface type and battery swapping service identification information of the acquired battery image. Optionally, the battery image feature dataset may include: first battery image feature data generated from semantic vectors of multiple regions of the battery image stored based on the image acquisition device identification information; second battery image feature data generated from semantic vectors of multiple regions of the battery image stored based on the battery identification information; third battery image feature data generated from semantic vectors of threatening anomaly regions; and fourth battery image feature data generated from semantic vectors of non-threatening anomaly regions. Threatening anomaly regions may include, but are not limited to, damaged areas, dented areas, foreign object areas, and other areas on the battery surface that pose a threat to the battery. Non-threatening abnormal areas may include, but are not limited to, areas on the battery surface such as water stains, oil stains, and other contaminants that do not pose a threat to the battery. Optionally, the battery anomaly detection results may include one or more of the following: whether the battery has abnormal areas, the number of abnormal areas, and a threat anomaly score corresponding to the abnormal areas.

[0060] Optionally, the anomaly detection module 220 can also be configured to send the battery anomaly detection results to the cloud storage module 250 for storage. Optionally, the anomaly detection module 220 can establish various forms of communication connections with the cloud storage module 250.

[0061] The judgment module 230 is configured to determine whether an abnormal region is a repeatedly detected abnormal region in response to a battery anomaly detection result indicating the presence of an abnormal region in the battery, and to send a battery anomaly detection result to the alarm module 240 in response to the determination that the abnormal region is not a repeatedly detected abnormal region. Optionally, the judgment module 230 can be configured to obtain historical battery anomaly detection results from the cloud storage module 250 and determine whether an abnormal region is a repeatedly detected abnormal region based on one or more of the following comparisons: a comparison between the number of abnormal regions and the number of abnormal regions detected in the previous detection; a comparison between the minimum repetition probability of the abnormal region and a probability threshold; and a comparison between the similarity between the semantic vector of the abnormal region and the semantic vector of the previously detected abnormal region and a similarity threshold. By determining whether the detected abnormal region is a repeatedly detected abnormal region, intelligent alarms can be achieved, reducing repetitive work and improving operation and maintenance efficiency while ensuring battery safety.

[0062] The alarm module 240 is configured to generate display data based on battery anomaly detection results. This display data is used to show non-repeated anomaly areas on the acquired battery image. Optionally, the anomaly areas displayed on the acquired battery image can be further displayed on the display interface of the battery maintenance center, the display interface of the battery swapping station, or the vehicle screen. Optionally, the alarm module 240 can also be configured to generate prompt information based on the battery anomaly detection results and send the prompt information to operators at the battery maintenance center, operators at the battery swapping station, or vehicle users.

[0063] The cloud storage module 250 may also be referred to as a cloud database in the context of this application. The data stored therein may include, but is not limited to, battery image feature datasets, raw image data, image size data, pixel position data corresponding to threat anomaly regions, threat anomaly regions and their labels, non-threat anomaly regions and their labels, battery anomaly detection results, acquired battery images, battery identification information, identification information of image acquisition devices, historical battery anomaly detection results, etc.

[0064] The anomaly detection training module 260 can be configured to update the anomaly detection module 220 based on one or more of the following data stored in the cloud storage module 250: original image data, image size data, pixel position data corresponding to the threat anomaly region, the threat anomaly region and its label, and the non-threat anomaly region and its label.

[0065] By establishing data interaction between the anomaly detection module 220, the cloud storage module 250, and the anomaly detection training module 260, the detection accuracy of the anomaly detection module 220 and the training efficiency of the anomaly detection training module 260 are improved. The anomaly detection module 220 not only performs online inference calculations based on the cloud storage module 250 and the anomaly detection training module 260, but also returns battery anomaly detection results to the cloud storage module 250, improving the robustness of the battery anomaly detection system 200. Furthermore, the battery anomaly detection system 200 according to one or more embodiments of this application possesses model self-iteration capabilities, exhibiting good system robustness after the cloud storage module 250 introduces more anomaly samples.

[0066] Understandable Figure 1 and Figure 2 The various modules in the battery anomaly detection system 100 and battery anomaly detection system 200 shown are merely illustrative and are not intended to depart from the spirit and scope of this application. Figure 1 and Figure 2The modules in the battery anomaly detection system 100 and battery anomaly detection system 200 shown in the figure can be merged or decomposed into other sub-modules, and this application does not limit this.

[0067] Figure 3 A schematic block diagram of a cloud database according to one or more embodiments of this application is shown.

[0068] like Figure 3 As shown, the cloud database 300 includes an anomaly target detection database 310, a first battery image feature database 320, a second battery image feature database 330, an anomaly region image classification database 340, a third battery image feature database 350, a fourth battery image feature database 360, an anomaly candidate database 370, and an alarm database 380.

[0069] The abnormal target detection database 310 stores raw image data, image size data, and pixel location data corresponding to the threat abnormal area.

[0070] The first battery image feature database 320 stores first battery image feature data generated from the semantic vectors of multiple regions of the battery image stored based on the identification information of the image acquisition device.

[0071] The second battery image feature database 330 stores second battery image feature data generated from the semantic vectors of multiple regions of the battery image stored based on battery identification information.

[0072] The abnormal region image classification database 340 stores threatening abnormal regions and their labels, as well as non-threatening abnormal regions and their labels.

[0073] The third battery image feature database 350 stores third battery image feature data generated from semantic vectors of threat-abnormal regions.

[0074] The fourth battery image feature database 360 ​​stores fourth battery image feature data generated from semantic vectors of non-threatening abnormal regions.

[0075] The anomaly candidate database 370 stores battery anomaly detection results, acquired battery images, battery identification information, and identification information of the image acquisition device. Optionally, the anomaly candidate database 370 can also receive manually input battery surface data.

[0076] The alarm database 380 stores historical battery anomaly detection results, which can be stored using battery identification information or image acquisition device identification information as an index.

[0077] Figure 4A schematic diagram of semantic vectors for generating a battery image of multiple regions according to one or more embodiments of this application is shown. Figure 4 The method shown can be used to process battery images stored based on the identification information of the image acquisition device and battery images stored based on the battery identification information to generate first battery image feature data and second battery image feature data, respectively.

[0078] like Figure 4 As shown, the battery image acquired by the image acquisition device is divided into multiple regions of equal size and each region is numbered. Then, each numbered region is encoded to generate a semantic vector for each region. For example, an image pre-trained encoder can be used to encode each numbered region to generate a one-dimensional semantic vector for each region.

[0079] Optionally, each semantic vector and its corresponding region code can be stored according to the identification information of the image acquisition device. Figure 3 The first battery image feature database 320 shown, and the storage of each semantic vector and its corresponding region code according to the battery's identification information. Figure 3 The second battery image feature database 330 shown is illustrated.

[0080] Figure 5 A schematic diagram illustrating the generation of semantic vectors for threat-anomalous regions and non-threat-anomalous regions according to one or more embodiments of this application is shown. Figure 5 The method shown can be used to process threatening and non-threatening threatening regions to generate third and fourth battery image feature data, respectively.

[0081] like Figure 5As shown, the size of each threat-anomaly region can be flattened to the same size region, and then the same size region is encoded to generate semantic vectors for each region. For example, an image pre-trained encoder can be used to encode the same size region to generate one-dimensional semantic vectors for each region. Since the number of threat-anomaly and non-threat-anomaly regions is large, which is not conducive to real-time detection by subsequent anomaly detection models, the generated semantic vectors can be clustered to select typical speech vectors. For example, a clustering algorithm can be used to treat all semantic vectors as a group of samples for clustering, where the distance between each semantic vector can be calculated using the cosine similarity between vectors, and a minimum number of samples is set for each cluster to obtain each cluster center sample and outlier sample. Optional clustering algorithms can be distance-based and minimum sample number-based clustering algorithms, including but not limited to DBSCAN, spectral clustering variants, etc. Finally, the semantic vectors corresponding to the cluster center samples are stored in a cloud database. Similarly, the same processing can be performed on non-threat-anomaly regions to obtain each cluster center sample and outlier sample, and the semantic vectors corresponding to the cluster center samples are stored in a cloud database.

[0082] Optionally, the semantic vectors corresponding to the cluster center samples of the threat anomaly region can be stored in [the appropriate storage location]. Figure 3 The third battery image feature database 350 shown, and the semantic vectors corresponding to the cluster center samples of non-threatening anomaly areas are stored in... Figure 3 The fourth battery image feature database 360 ​​is shown.

[0083] Figure 6 A schematic diagram of an anomaly detection module according to one or more embodiments of this application is shown.

[0084] like Figure 6 As shown, the anomaly detection module 620 may include an anomaly warning unit 6201 and an anomaly diagnosis unit 6202.

[0085] The anomaly warning unit 6201 can be configured to generate an anomaly warning area based at least on the acquired battery image, first battery image feature data, and second battery image feature data. The anomaly diagnosis unit 6202 can be configured to generate a threat anomaly score for the anomaly warning area based at least on the anomaly warning area generated by the anomaly warning unit 6201, third battery image feature data, and fourth battery image feature data.

[0086] The anomaly warning unit 6201 may include: a target detection subunit configured to detect an acquired battery image and generate a first anomaly warning region in response to detecting an abnormal target in the acquired battery image that is greater than an anomaly warning score threshold; a first comparison subunit configured to process the acquired battery image to obtain semantic vectors of multiple regions of the acquired battery image and compare the semantic vectors of the multiple regions of the acquired battery image with first battery image feature data to generate a second anomaly warning region; and a second comparison subunit configured to process the acquired battery image to obtain semantic vectors of multiple regions of the acquired battery image and compare the semantic vectors of the multiple regions of the acquired battery image with second battery image feature data to generate a third anomaly warning region. Exemplarily, the first comparison subunit and the second comparison subunit may be configured to utilize... Figure 4 The process described above is used to process the acquired battery images to obtain semantic vectors for multiple regions of the acquired battery images, and to calculate the cosine similarity between the semantic vector corresponding to the region number and the corresponding first battery image feature data and second battery image feature data. When the cosine similarity of some regions is lower than the threshold similarity, the region is regarded as an abnormal warning region.

[0087] Optionally, the target detection subunit, the first comparison subunit, and the second comparison subunit of the anomaly warning unit 6201 can operate in parallel, and the anomaly warning unit 6201 can be further configured to generate an anomaly warning region based on a first anomaly warning region, a second anomaly warning region, and a third anomaly warning region. For example, the anomaly warning unit 6201 can be further configured to superimpose the first anomaly warning region, the second anomaly warning region, and the third anomaly warning region to generate an anomaly warning region.

[0088] Optionally, the object detection subunit can be implemented using an optimal anomaly warning detection model trained based on an object detection supervised algorithm. This model can generate a first anomaly warning region when an anomaly target greater than the anomaly warning score threshold is detected, and can optionally perform non-maximum suppression before generating the first anomaly warning region. Optionally, the training data for the anomaly warning detection model can be taken from the anomaly object detection database 310 in the cloud database 300, and the object detection algorithms that can be used include, but are not limited to, Faster R-CNN, YOLOv5, etc.

[0089] Optionally, the first and second comparison sub-units can be updated based on data in the cloud database to improve the accuracy of anomaly detection by the anomaly detection module 620. Optionally, when the number of new data records in the anomaly detection database 310 in the cloud database 300 exceeds a certain threshold, the first and second comparison sub-units can be automatically updated based on data in the anomaly detection database 310. In one embodiment, the first and second comparison sub-models can be trained using, for example, 80% of the data in the anomaly detection database 310 as a training set, and the remaining 20% ​​of the data in the anomaly detection database 310 as a validation set to calculate evaluation metrics such as accuracy, recall, and precision of the first and second comparison sub-models. If the evaluation metrics of the trained first and second comparison sub-models are better than the current first and second comparison models, the parameters of the trained first and second comparison models are used to update the first and second comparison sub-units.

[0090] The anomaly diagnosis unit 6202 may include: a classification subunit configured to process anomaly warning regions to generate anomaly scores for the anomaly warning regions; a first comparison subunit configured to process anomaly warning regions to obtain semantic vectors of the anomaly warning regions and compare the semantic vectors of the anomaly warning regions with third battery image feature data to generate a first similarity score for the anomaly warning regions; and a second comparison subunit configured to process anomaly warning regions to obtain semantic vectors of the anomaly warning regions and compare the semantic vectors of the anomaly warning regions with fourth battery image feature data to generate a second similarity score for the anomaly warning regions. Optionally, the classification subunit can be implemented using an optimal classification model trained based on an image classification supervised algorithm, where the processing result for each anomaly warning region is an anomaly score with a softmax or sigmoid function output value between 0 and 1. Optionally, the training data for the optimal classification model can be taken from the anomaly region image classification database 340 in the cloud database 300, and the image algorithms that can be used may include, but are not limited to, VGGNet, ResNet, etc. Optionally, the first comparison subunit can use a pre-trained encoder to encode the anomaly warning region to obtain the semantic vector of the anomaly warning region, calculate the cosine similarity between the semantic vector of the anomaly warning region and the third battery image feature data, normalize the cosine similarity to between 0 and 1, and finally take the maximum value of the normalized cosine similarity as the first similarity score for the anomaly warning region. Optionally, the second comparison subunit can use a pre-trained encoder to encode the anomaly warning region to obtain the semantic vector of the anomaly warning region, calculate the cosine similarity between the semantic vector of the anomaly warning region and the fourth battery image feature data, normalize the cosine similarity to between 0 and 1, and finally take the maximum value of the normalized cosine similarity as the second similarity score ψ for the anomaly warning region. k Optionally, the comparison operation between the first comparison subunit and the second comparison subunit can be performed using matrix operations to improve computational efficiency.

[0091] Optionally, the first and second comparison sub-units can be updated based on data in the cloud database to improve the accuracy of anomaly detection by the anomaly detection module 620. Optionally, when the number of new data records in the anomaly region image classification database 340 in the cloud database 300 exceeds a specific threshold, the first and second comparison sub-units can be automatically updated based on data in the anomaly region image classification database 340. In one embodiment, the first and second comparison sub-models can be trained using, for example, 80% of the data in the anomaly region image classification database 340 as a training set, and the remaining 20% ​​of the data in the anomaly region image classification database 340 can be used as a validation set to calculate evaluation metrics such as accuracy, recall, and precision of the first and second comparison sub-models. If the evaluation metrics of the trained first and second comparison sub-models are better than the current first and second comparison sub-models, then the first and second comparison sub-units are updated using the parameters of the trained first and second comparison sub-models.

[0092] Optionally, the anomaly diagnosis unit 6202 can be further configured to pair anomaly scores ψ i First similarity score ψ j Second similarity score ψ k Weighted processing is performed to generate a threat anomaly score for the anomaly warning area, and an anomaly area indicating the presence of the battery as a result of the battery anomaly detection is generated in response to the threat anomaly score for the anomaly warning area being greater than a threat anomaly score threshold. For example, the anomaly diagnosis unit 6202 can use the following formula (1) to calculate the anomaly score ψ. i First similarity score ψ j Second similarity score ψ k Weighted processing is performed to generate a threat anomaly score Φ for the aforementioned anomaly warning area:

[0093] Φ=w i ψ i +w j ψ j +w k (1-ψ k ) Formula (1)

[0094] Where w i w j w k This is the weighting factor for the corresponding score.

[0095] By decoupling the anomaly detection module 620 into an anomaly warning unit 6201 and an anomaly diagnosis unit 6202, it can accurately distinguish between threatening and non-threatening anomaly areas. Specifically, it can accurately identify whether detected abnormal areas on the battery surface are damage, dents, water stains, oil stains, etc., thereby improving alarm accuracy. Furthermore, the anomaly detection module 620 integrates supervised learning and similarity comparison methods instead of a single supervised learning method. This allows it to maintain high detection accuracy even with limited battery surface data samples. It can detect not only known types of battery surface anomalies but also provide warnings for unknown types, solving the problem of insufficient battery surface anomaly samples.

[0096] Figure 7 A schematic diagram of a comparison subunit of an anomaly warning unit according to one or more embodiments of this application is shown. Figure 7 The working principle shown can be applied to the reference. Figure 6 The first comparison subunit and the second comparison subunit of the abnormal early warning unit 6201.

[0097] like Figure 7 As shown, the acquired battery image is divided into multiple regions of equal size and each region is numbered. Then, each numbered region is encoded to generate semantic vectors for multiple regions of the acquired battery image. For example, an image pre-trained encoder can be used to encode each numbered region to generate a one-dimensional semantic vector for each region. Next, the cosine similarity between the semantic vector corresponding to the region number and the corresponding first battery image feature data, and the cosine similarity between the semantic vector corresponding to the region number and the corresponding second battery image feature data, are calculated. When the cosine similarity of some numbered regions is lower than a threshold similarity, that region is designated as an abnormal warning region.

[0098] Figure 8 A schematic diagram of a comparison subunit of an anomaly diagnosis unit according to one or more embodiments of this application is shown. Figure 8 The working principle shown can be applied to the reference. Figure 6 The abnormality diagnosis unit 6202 includes a first comparison subunit and a second comparison subunit.

[0099] like Figure 8 As shown, a pre-trained encoder can be used to encode the anomaly warning region to obtain the semantic vector of the anomaly warning region, and the cosine similarity between the semantic vector of the anomaly warning region and the third battery image feature data / fourth battery image feature data can be calculated and normalized to between 0 and 1. Finally, the maximum value of the normalized cosine similarity is taken as the similarity score for the anomaly warning region.

[0100] Figure 9 A flowchart of a battery anomaly detection method according to one or more embodiments of this application is shown.

[0101] like Figure 9 As shown, in step S901, the battery image, battery identification information, and identification information of the image acquisition device are acquired by the image acquisition device.

[0102] In step S903, the anomaly detection module obtains a battery image feature dataset from the cloud database based on the battery identification information received from the communication module and the identification information of the image acquisition device, and generates a battery anomaly detection result based at least on the acquired battery image and the battery image feature dataset.

[0103] In step S905, in response to the battery anomaly detection result indicating the existence of an abnormal region in the battery, the determination module determines whether the abnormal region is a repeatedly detected abnormal region, and in response to determining that the abnormal region is not a repeatedly detected abnormal region, sends the battery anomaly detection result. Optionally, the determination module determines whether the abnormal region is a repeatedly detected abnormal region based on one or more of the following comparisons: a comparison between the number of abnormal regions and the number of abnormal regions detected previously; a comparison between the minimum repetition probability of the abnormal region and a probability threshold; and a comparison between the similarity between the semantic vector of the abnormal region and the semantic vector of the previously detected abnormal region and a similarity threshold.

[0104] Optionally, the battery image feature dataset includes: first battery image feature data generated from semantic vectors of multiple regions of the battery image stored based on the identification information of the image acquisition device; second battery image feature data generated from semantic vectors of multiple regions of the battery image stored based on the battery identification information; third battery image feature data generated from semantic vectors of threatening abnormal regions on the battery surface; and fourth battery image feature data generated from semantic vectors of non-threatening abnormal regions on the battery surface.

[0105] Optionally, the battery anomaly detection method further includes sending one or more of the following to the cloud database for storage: the battery anomaly detection result, the acquired battery image, the battery identification information, and the identification information of the image acquisition device. Optionally, the battery anomaly detection method further includes generating display data based on the battery anomaly detection result, the display data being used to display the abnormal area on the acquired battery image. Optionally, the battery anomaly detection method further includes updating the anomaly detection module based on one or more of the following data from the cloud database: original image data, image size data, pixel position data corresponding to the threat anomaly area, the threat anomaly area and its label, and the non-threat anomaly area and its label.

[0106] The battery anomaly detection method according to one aspect of this application can accurately identify abnormal areas in the battery through an anomaly detection module, and can prevent repeated detection of the same abnormal area through a judgment module. This improves the accuracy of battery anomaly detection, ensuring battery safety and enhancing operational efficiency.

[0107] Figure 10 A flowchart illustrating a method for determining whether an abnormal region is a repeatedly detected abnormal region according to one or more embodiments of this application is shown.

[0108] like Figure 10 As shown, in step S1001, the number of abnormal regions detected this time is compared with the number of abnormal regions detected last time to determine whether the number of abnormal regions detected this time is greater than the number of abnormal regions detected last time. If it is greater, the abnormal regions detected this time are determined to be non-repeated abnormal regions; otherwise, proceed to step S1003.

[0109] In step S1003, the minimum repetition probability of the abnormal region is compared with a probability threshold to determine whether the minimum repetition probability of the abnormal region is less than the probability threshold. If it is less, the abnormal region detected this time is determined to be a non-repetitive abnormal region; otherwise, proceed to step S1005. Optionally, in step S1003, the intersection-union ratio (IUR) between the abnormal region detected this time and the abnormal region detected last time can be calculated, and the maximum IUR is taken as the repetition probability of the abnormal region. The corresponding abnormal region detected last time is taken as the region that matches the abnormal region detected this time. If the minimum repetition probability of each abnormal region detected this time is less than the probability threshold, the abnormal region detected this time is determined to be a non-repetitive abnormal region; otherwise, proceed to step S1005.

[0110] In step S1005, the similarity between the semantic vector of the abnormal region and the semantic vector of the previously detected abnormal region is compared with a similarity threshold to determine whether the similarity between the semantic vector of the abnormal region and the semantic vector of the previously detected abnormal region is greater than the similarity threshold. If it is greater, the abnormal region detected this time is determined to be a repeatedly detected abnormal region; otherwise, the abnormal region detected this time is determined to be a non-repeated detected abnormal region. Optionally, in step S1005, a pre-trained encoder can be used to encode the abnormal region detected this time and the matched previously detected abnormal region into semantic vectors, and the cosine similarity between the semantic vector of the abnormal region detected this time and the semantic vector of the matched previously detected abnormal region is calculated, and the cosine similarity is compared with a similarity threshold. If the cosine similarity of each abnormal region detected this time is greater than the similarity threshold, the abnormal region detected this time is determined to be a repeatedly detected abnormal region; otherwise, the abnormal region detected this time is determined to be a non-repeated detected abnormal region.

[0111] Figure 11 A block diagram of a computer device according to one or more embodiments of this application is shown. As shown in Figure 11, the computer device 1100 includes a memory 1110, a processor 1120, and a computer program 1130 stored in the memory 1110 and executable on the processor 1120. When the processor 1120 executes the computer program 1130, it implements the various steps of the battery anomaly detection method as described above.

[0112] Alternatively, this application can also be implemented as a computer storage medium storing a program for causing a computer to execute a battery anomaly detection method according to one aspect of this application.

[0113] As a computer storage medium, various types of computer storage media can be used, such as disks (e.g., hard disks, optical disks, etc.), cards (e.g., memory cards, optical cards, etc.), semiconductor memory (e.g., ROM, non-volatile memory, etc.), and tapes (e.g., magnetic tape, cassette tape, etc.).

[0114] Additionally, as described above, this application can also be implemented as a battery swapping station, comprising an image acquisition device configured to acquire battery images; a communication device configured to transmit the acquired battery images to a battery anomaly detection system according to one aspect of this application and to receive display data from the battery anomaly detection system; and a display device configured to display a battery anomaly area on the acquired battery images based on the display data.

[0115] Where applicable, the various embodiments provided in this application may be implemented using hardware, software, or a combination of hardware and software. Furthermore, where applicable, without departing from the scope of this application, the various hardware and / or software components described herein may be combined into composite components comprising software, hardware, and / or both. Where applicable, without departing from the scope of this application, the various hardware and / or software components described herein may be divided into sub-components comprising software, hardware, or both. Additionally, where applicable, it is contemplated that software components may be implemented as hardware components, and vice versa.

[0116] The software (such as program code and / or data) according to this application can be stored on one or more computer storage media. It is also contemplated that the software identified herein can be implemented using one or more networked and / or otherwise general-purpose or special-purpose computers and / or computer systems. Where applicable, the order of the various steps described herein can be changed, combined into compound steps, and / or divided into sub-steps to provide the features described herein.

[0117] The embodiments and examples presented herein are provided to best illustrate embodiments of this application and its particular applications, thereby enabling those skilled in the art to implement and use this application. However, those skilled in the art will understand that the above description and examples are provided for ease of illustration and example only. The descriptions presented are not intended to cover all aspects of this application or to limit this application to the precise forms disclosed.

Claims

1. A battery anomaly detection system, characterized in that, The system includes: A communication module configured to acquire battery images, battery identification information, and identification information of the image acquisition device acquired by the image acquisition device; An anomaly detection module is configured to retrieve a battery image feature dataset from a cloud database based on the battery identification information received from the communication module and the identification information of the image acquisition device, and to generate a battery anomaly detection result based at least on the acquired battery image and the battery image feature dataset; and The judgment module is configured to determine whether the abnormal region is a repeatedly detected abnormal region in response to the battery abnormality detection result indicating that the battery has an abnormal region, and to send the battery abnormality detection result in response to the determination that the abnormal region is not a repeatedly detected abnormal region. The battery image feature dataset includes: first battery image feature data generated from semantic vectors of multiple regions of the battery image stored based on the identification information of the image acquisition device; second battery image feature data generated from semantic vectors of multiple regions of the battery image stored based on the battery identification information; third battery image feature data generated from semantic vectors of threat-anomalous regions; and fourth battery image feature data generated from semantic vectors of non-threat-anomalous regions. The anomaly detection module includes: An anomaly warning unit is configured to generate an anomaly warning area based at least on the acquired battery image, the first battery image feature data, and the second battery image feature data; and An anomaly diagnosis unit is configured to generate a threat anomaly score for the anomaly warning area based at least on the anomaly warning area, the third battery image feature data, and the fourth battery image feature data. The abnormality diagnosis unit includes: A classification subunit is configured to process the anomaly warning region to generate an anomaly score for the anomaly warning region; The first comparison subunit is configured to process the anomaly warning region to obtain the semantic vector of the anomaly warning region and compare the semantic vector of the anomaly warning region with the third battery image feature data to generate a first similarity score for the anomaly warning region. The second comparison subunit is configured to process the anomaly warning region to obtain a semantic vector of the anomaly warning region and compare the semantic vector of the anomaly warning region with the fourth battery image feature data to generate a second similarity score for the anomaly warning region.

2. The system according to claim 1, wherein the anomaly detection module is further configured to: One or more of the battery anomaly detection results, the acquired battery image, the battery identification information, and the identification information of the image acquisition device are sent to the cloud database for storage.

3. The system according to claim 1, wherein the cloud database stores one or more of the following data: the battery image feature dataset, original image data, image size data, pixel position data corresponding to the threat anomaly region, threat anomaly region and its label, non-threat anomaly region and its label, battery anomaly detection result, the acquired battery image, battery identification information, identification information of the image acquisition device, and historical battery anomaly detection results.

4. The system according to claim 1, wherein the anomaly warning unit comprises: A target detection subunit is configured to detect the acquired battery image and generate a first abnormal warning area in response to the detection of an abnormal target in the acquired battery image that is greater than an abnormal warning score threshold. The first comparison subunit is configured to process the acquired battery image to obtain semantic vectors of multiple regions of the acquired battery image and compare the semantic vectors of the multiple regions of the acquired battery image with the first battery image feature data to generate a second abnormal warning region. as well as The second comparison subunit is configured to process the acquired battery image to obtain semantic vectors of multiple regions of the acquired battery image and compare the semantic vectors of the multiple regions of the acquired battery image with the second battery image feature data to generate a third abnormal warning region.

5. The system according to claim 4, wherein the anomaly warning unit is further configured to generate the anomaly warning region based on the first anomaly warning region, the second anomaly warning region, and the third anomaly warning region.

6. The system according to claim 1, wherein the anomaly diagnosis unit is further configured to: The anomaly score, the first similarity score, and the second similarity score are weighted to generate a threat anomaly score for the anomaly warning area; and In response to a threat anomaly score greater than a threat anomaly score threshold for the anomaly warning area, the anomaly warning area is used as the anomaly area indicating the presence of the battery as indicated by the battery anomaly detection result.

7. The system according to claim 6, wherein the battery anomaly detection result includes one or more of the following: whether the battery has the abnormal region, the number of the abnormal regions, and the threat anomaly score corresponding to the abnormal region.

8. The system of claim 1, wherein the system further comprises: An alarm module is configured to generate display data based on the battery anomaly detection results. The display data is used to display the abnormal area on the acquired battery image.

9. The system of claim 1, wherein the system further comprises: An anomaly detection training module is configured to update the anomaly detection module based on one or more of the following data in the cloud database: original image data, image size data, pixel position data corresponding to threat anomaly regions, threat anomaly regions and their labels, and non-threat anomaly regions and their labels.

10. The system according to claim 1, wherein the determining module is further configured to determine whether the abnormal region is a repeatedly detected abnormal region based on one or more of the following comparisons: A comparison between the number of abnormal regions and the number of abnormal regions detected in the previous test; The comparison between the minimum repetition probability of the abnormal region and the probability threshold; and The comparison is made between the similarity between the semantic vector of the abnormal region and the semantic vector of the previously detected abnormal region and a similarity threshold.

11. The system of claim 1, wherein the image acquisition device is located at the battery swapping station and configured to acquire battery images during battery swapping operations.

12. A method for detecting battery anomalies, characterized in that, The method includes: Acquire battery images, battery identification information, and identification information of the image acquisition device acquired by the image acquisition device; The anomaly detection module retrieves a battery image feature dataset from a cloud database based on the battery identification information received from the communication module and the identification information of the image acquisition device, and generates a battery anomaly detection result based at least on the acquired battery images and the battery image feature dataset; and In response to the battery anomaly detection result indicating the presence of an abnormal region in the battery, the determination module determines whether the abnormal region is a repeatedly detected abnormal region, and in response to determining that the abnormal region is not a repeatedly detected abnormal region, the battery anomaly detection result is sent. The battery image feature dataset includes: first battery image feature data generated from semantic vectors of multiple regions of the battery image stored based on the identification information of the image acquisition device; second battery image feature data generated from semantic vectors of multiple regions of the battery image stored based on the battery identification information; third battery image feature data generated from semantic vectors of threat-anomalous regions; and fourth battery image feature data generated from semantic vectors of non-threat-anomalous regions. The method further includes the anomaly detection module: An abnormal warning area is generated based at least on the acquired battery image, the first battery image feature data, and the second battery image feature data; and A threat anomaly score is generated for the anomaly warning area based at least on the anomaly warning area, the third battery image feature data, and the fourth battery image feature data. The method further includes: Process the anomaly warning area to generate an anomaly score for the anomaly warning area; The abnormal warning region is processed to obtain the semantic vector of the abnormal warning region, and the semantic vector of the abnormal warning region is compared with the third battery image feature data to generate a first similarity score for the abnormal warning region. The abnormal warning region is processed to obtain the semantic vector of the abnormal warning region, and the semantic vector of the abnormal warning region is compared with the fourth battery image feature data to generate a second similarity score for the abnormal warning region.

13. The method of claim 12, wherein the method further comprises: One or more of the battery anomaly detection results, the acquired battery image, the battery identification information, and the identification information of the image acquisition device are sent to the cloud database for storage.

14. The method according to claim 12, wherein the cloud database stores one or more of the following data: the battery image feature dataset, original image data, image size data, pixel position data corresponding to the threat anomaly region, threat anomaly region and its label, non-threat anomaly region and its label, battery anomaly detection result, the acquired battery image, battery identification information, identification information of the image acquisition device, and historical battery anomaly detection results.

15. The method of claim 12, wherein the method further comprises: Display data is generated based on the battery anomaly detection results, and the display data is used to display the abnormal area on the acquired battery image.

16. The method of claim 12, wherein the method further comprises: The anomaly detection module is updated based on one or more of the following data in the cloud database: original image data, image size data, pixel position data corresponding to the threat anomaly region, threat anomaly region and its label, non-threat anomaly region and its label.

17. The method of claim 12, wherein determining whether the abnormal region is a repeatedly detected abnormal region by the determining module includes determining whether the abnormal region is a repeatedly detected abnormal region based on one or more of the following comparisons: A comparison between the number of abnormal regions and the number of abnormal regions detected in the previous test; The comparison between the minimum repetition probability of the abnormal region and the probability threshold; and The comparison is made between the similarity between the semantic vector of the abnormal region and the semantic vector of the previously detected abnormal region and a similarity threshold.

18. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program to achieve: The battery anomaly detection method according to any one of claims 12-17.

19. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes instructions that, when executed, perform the battery anomaly detection method according to any one of claims 12-17.

20. A battery swapping station, characterized in that, The battery swapping station includes: An image acquisition device configured to acquire images of a battery; A communication device configured to transmit acquired battery images to a battery anomaly detection system according to any one of claims 1-11 and receive display data from the battery anomaly detection system; and A display device configured to display abnormal battery areas on the acquired battery image based on the display data.

Images