Data center inspection system, battery anomaly identification method, device, and storage medium

Abstract

Embodiments of the present application provide a data center inspection system, a battery anomaly identification method, equipment and a storage medium. In the embodiments of the present application, a battery array image of a data center collected by a self-moving device can be obtained; and battery features of the battery array image are extracted to obtain battery features reflected by the battery array image; then, according to the battery features reflected by the battery array image, an abnormal battery in a target battery included in the battery array image and a target abnormal type of the abnormal battery can be identified, so as to realize automatic identification of the abnormal battery and the battery abnormal type of the abnormal battery, which helps to improve the battery anomaly identification efficiency compared with the manual inspection mode. On the other hand, compared with the manual inspection mode, the battery anomaly missed detection and misidentification caused by insufficient experience or negligence of the manual inspection can be avoided, which helps to reduce the probability of the battery anomaly missed detection and the probability of the misidentification.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a data center inspection system, a battery anomaly identification method, device, and storage medium. Background Technology

[0002] Data centers consist of many large-scale cluster systems, including not only computing cluster systems but also other supporting equipment such as communication equipment, storage devices, and power supply systems. The power supply system provides electricity to the data center, ensuring its normal operation.

[0003] Batteries are one of the core facilities of a data center, serving as the last line of defense for ensuring power supply. To promptly detect and repair any anomalies in the batteries, data center maintenance personnel need to inspect them multiple times daily. The current practice is manual inspection, which involves visually inspecting each battery cell to identify any abnormalities. Manual inspection has the following drawbacks: it is extremely inefficient and prone to missing anomalies. Summary of the Invention

[0004] This application provides a data center inspection system, a battery anomaly identification method, a device, and a storage medium to improve the efficiency of data center battery inspection.

[0005] This application provides a data center inspection system, including: an autonomous mobile device and a server device; the autonomous mobile device is equipped with an image acquisition device;

[0006] The autonomous mobile device is used to move along the battery array in the data center, and during the movement along the battery array, controls the image acquisition device to acquire images of the battery array; and provides the battery array images to the server device.

[0007] The server-side device is configured to: extract battery features from the battery array image to determine the battery features of the battery array image; and, based on the battery features, identify anomalies in the target batteries contained in the battery array image to determine the abnormal batteries in the target batteries and the target anomaly type of the abnormal batteries.

[0008] This application also provides a battery anomaly identification method, including:

[0009] Acquire images of the battery array as an autonomous mobile device moves along the battery array in a data center;

[0010] Battery features are extracted from the battery array image to determine the battery features of the battery array image;

[0011] Based on the battery characteristics, anomaly identification is performed on the target batteries contained in the battery array image to determine the abnormal batteries in the target batteries and the target anomaly type of the abnormal batteries.

[0012] This application embodiment also provides an autonomous mobile device, including: a mechanical body; an image acquisition device, a memory, and a processor are installed on the mechanical body; the memory is used to store computer programs;

[0013] The image acquisition device is used to acquire images of the battery array during the movement of the autonomous mobile device along the battery array in the data center.

[0014] The processor is coupled to the memory and is used to execute the computer program to perform the steps in the above-described battery anomaly identification method.

[0015] This application embodiment also provides a computing device, including: a memory, a processor, and a communication component; the memory is used to store computer programs;

[0016] The processor is coupled to the memory and the communication component to execute the computer program for performing the steps in the above-described battery anomaly identification method.

[0017] This application also provides a computer-readable storage medium storing computer instructions, which, when executed by one or more processors, cause the one or more processors to perform the steps in the above-described battery anomaly identification method.

[0018] In this embodiment, battery array images of a data center collected by an autonomous mobile device can be acquired. Battery features are extracted from the battery array images to obtain the battery characteristics reflected in the images. Then, based on the battery characteristics reflected in the images, anomaly identification is performed on the target batteries within the images to determine the target anomaly type. This achieves automatic identification of battery anomaly types, which helps improve the efficiency of battery anomaly identification compared to manual inspection. Furthermore, compared to manual inspection, it avoids missed or misidentified battery anomalies due to insufficient human experience or negligence, helping to reduce the probability of missed or misidentified battery anomalies. Attached Figure Description

[0019] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments of this application and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0020] Figure 1 This is a schematic diagram of the data center inspection system provided in an embodiment of this application;

[0021] Figure 2 This is a schematic diagram of the inspection process of the data center inspection system provided in the embodiments of this application;

[0022] Figure 3 This is a flowchart illustrating the battery anomaly identification method provided in an embodiment of this application.

[0023] Figure 4 This is a schematic diagram of the structure of a computing device provided in an embodiment of this application;

[0024] Figure 5 A schematic diagram of the structure of the autonomous mobile device provided in the embodiments of this application. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0026] To address the drawbacks of manual inspection in data centers—namely, extremely low efficiency and a high risk of missed anomaly detection—this application addresses several issues. In some embodiments, battery array images of the data center are acquired via autonomous mobile devices. Battery features are extracted from these images to reveal their characteristics. Based on these features, anomaly detection is performed on target batteries within the battery array image to identify the abnormal batteries and their specific anomalies. This automatic identification of battery anomaly types significantly improves efficiency compared to manual inspection. Furthermore, this method avoids missed or misidentified battery anomalies due to human inexperience or negligence, reducing the probability of both.

[0027] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0028] It should be noted that the same reference numerals denote the same object in the following figures and embodiments. Therefore, once an object is defined in one figure or embodiment, it does not need to be discussed further in subsequent figures and embodiments.

[0029] Figure 1 This is a schematic diagram of the data center inspection system provided in an embodiment of this application. Figure 1As shown, the data center inspection system may include: autonomous mobile device 10 and server device 20.

[0030] In this embodiment, the autonomous mobile device 10 can move autonomously and complete some tasks based on its autonomous movement. In this embodiment, the specific implementation of the autonomous mobile device 10 is not limited. The autonomous mobile device 10 can be implemented as a robot or a drone, etc. The robot's appearance can be humanoid, animal-shaped, vehicle-shaped, or puppet-shaped, etc. In this embodiment, as... Figure 1 As shown, an image acquisition device 101 (corresponding to) is installed on the autonomous mobile device 10. Figure 2 (Image acquisition in process 1). In this embodiment, the image acquisition device can be any device with image acquisition capabilities, such as a camera, video recorder, etc. The images acquired by the image acquisition device can be independent frames or video frames from a video.

[0031] In this embodiment, the autonomous mobile device 10 can move within the data center. In this embodiment, the autonomous mobile device 10 refers to a device with an independent power system. The autonomous mobile device 10 can move using its own power system. The power system may include drive wheels, a drive motor, and a transmission device. The autonomous mobile device 10 can move automatically using its own power system according to an inspection route. For example, the autonomous mobile device 10 can automatically plan an inspection route and move automatically according to that route. Specific implementation methods for planning inspection routes will be described in the following embodiments and will not be repeated here. Of course, the autonomous mobile device 10 can also be controlled by a user or other devices to move within the data center. For example, a computing device (such as server device 20) can control the autonomous mobile device 10 to move within the data center. Optionally, the computing device can send an inspection route to the autonomous mobile device 10 and control the autonomous mobile device 10 to move within the data center according to the inspection route. For another example, a user can control the autonomous mobile device 10 to move within the data center via a mobile phone, remote control, or other terminal. The autonomous mobile device 10 can move using its own power system in response to terminal control signals.

[0032] Specifically, the autonomous mobile device 10 moves within the battery compartment of the data center (corresponding to...). Figure 2 (Autonomous movement and positioning in process 1). The data center's battery room stores batteries in the form of battery arrays. A battery array refers to storing batteries in an array format. For example... Figure 1As shown, one rack can correspond to one or more battery arrays, but is not limited to this. "Multiple" refers to two or more. The batteries can be electric batteries and / or uninterruptible power supplies (UPS), etc. In this embodiment, the batteries are mainly used to ensure uninterrupted power supply to the data center, that is, to continue providing power to the data center when the AC or DC power supply (such as mains power) fails or malfunctions, thus preventing power outages to the data center equipment. Therefore, the quality of the batteries is crucial to ensuring the power supply and stable operation of the data center. To promptly detect and repair various abnormalities in the batteries, it is necessary to inspect the data center's battery arrays. In this embodiment, the data center inspection system mainly refers to a system for inspecting the data center's battery arrays, and can be called a data center battery inspection system.

[0033] In this embodiment, to improve the inspection effect of the battery array, an autonomous mobile device 10 can inspect the battery array in the data center. For batteries, battery abnormalities can be reflected to some extent on their external surface. These abnormalities include, but are not limited to, leakage, bulging, corrosion, or acid creep. Therefore, in this embodiment, the autonomous mobile device 10 can move along the battery array in the data center, and during this movement, the image acquisition device 101 is controlled to acquire images of the battery array, i.e., battery array images. In this embodiment, the specific route taken by the autonomous mobile device 10 along the battery array in the data center is not limited.

[0034] In some embodiments, an inspection route can be pre-planned for the autonomous mobile device 10; and the inspection route is pre-set in the autonomous mobile device 10. In this way, the autonomous mobile device 10 can move along the battery array in the data center according to the set inspection route. In this application embodiment, the specific implementation method and execution device for pre-generating the inspection route are not limited. Optionally, the location distribution of the battery array in the environmental map of the data center can be determined; and based on the location distribution of the battery array in the environmental map, an inspection route parallel to the battery array (corresponding to) can be planned for the autonomous mobile device. Figure 2 (The inspection route is determined in process 2). This inspection route can then be pre-set in the autonomous mobile device 10. The device that plans the inspection route can be the autonomous mobile device 10, the server device 20, or any other computing device.

[0035] Optionally, when planning an inspection route parallel to the battery array for the autonomous mobile device based on the battery array's location distribution on the environmental map, the path planning algorithm used includes, but is not limited to, the D* algorithm, A* algorithm, genetic algorithm, ion swarm algorithm, or ant colony algorithm. Among these, the A* algorithm is a heuristic search algorithm, and the D* algorithm is a reverse heuristic search algorithm.

[0036] Heuristic search algorithms can establish heuristic search rules during the search process to measure the distance relationship between the search location and the target location, prioritizing the direction of the target point's location in the search, thus improving search performance. In this embodiment, the target location is the inspection termination point; the search location refers to the current location of the autonomous mobile device. The main idea of ​​the heuristic search algorithm is to introduce an estimation function f(x) for the current search node x. The estimation function f(x) can be expressed as:

[0037] f(x)=g(x)+h(x) (1).

[0038] In equation (1) above, g(x) refers to the actual distance from the starting point to the current search node x; h(x) refers to the minimum distance estimate from the current search node x to the target location. In this embodiment, the starting point can be the starting point of the autonomous mobile device, and the ending point is the inspection ending point of the autonomous mobile device. The starting point and the ending point can be the same point. h(x) can be measured using Euclidean distance, cosine distance, or Manhattan distance. Based on equation (1) above, when the autonomous mobile device uses the heuristic search algorithm for route planning, it can calculate the f value of each child node of the starting position, i.e., the value of equation (1) above, starting from the starting position. From the child nodes of the starting position, the child node with the smallest f value is selected as the next point of the starting position; then, in the same way, the child nodes of the next point of the starting position are determined, and the process is repeated until the next child node is the inspection ending point. The route formed by each determined child node, together with the starting point and the specified area location, is the planned inspection route.

[0039] The reverse incremental search algorithm is an improvement on the heuristic search algorithm. "Reverse" refers to searching step-by-step from the end point of the inspection back to the starting point; "incremental search" refers to calculating the distance metric D(x) for each node x during the search process. The distance metric D(x) can be expressed as:

[0040] D(x)=D(y)+D(y,x) (2).

[0041] In equation (2), D(y) represents the actual distance from node y to the target location; D(y, x) represents the actual distance from node x to node y. For the reverse incremental search algorithm, the f-value of each child node in the specified area can be calculated, i.e., the value of equation (1) above. From the child nodes of the inspection termination point, the child node with the smallest f-value is selected as the next point of the inspection termination point. Then, in the same way, the child nodes of the next point of the inspection termination point are determined, and the process is repeated until the next child node is the starting point. The route formed by each determined child node, together with the starting position and the specified area position, is the planned inspection route.

[0042] The autonomous mobile device can move along the initial inspection route. During the movement, if an obstacle is encountered that prevents the autonomous mobile device from continuing along the initial navigation route, the autonomous mobile device can use a reverse incremental search algorithm to replan the inspection route based on the distance metric information of each point not traversed in the initial navigation path. Specifically, assuming the autonomous mobile device discovers an obstacle at position x that is the next node of node x on the initial navigation route, the autonomous mobile device first calculates the actual distance from the current position x to the inspection termination point, and calculates the distance between node x and its new child node y. From the new child nodes y of node x, the child node with the smallest f value is selected as the new next point of node x. This process is iterated repeatedly until the next child node becomes the starting point, thus obtaining a new inspection route.

[0043] The inspection route planning method shown above is merely an illustrative example and does not constitute a limitation.

[0044] In other embodiments, the autonomous mobile device 10 can also perform autonomous positioning and navigation during movement (corresponding to...). Figure 2 In process 1, the autonomous mobile device 10 can autonomously move and locate itself, and plan its inspection route while moving. The autonomous mobile device 10 can achieve autonomous positioning and navigation inspection through Simultaneous Localization and Mapping (SLAM) technology to obtain the inspection route.

[0045] In this embodiment, as the autonomous mobile device 10 moves along the battery array according to a set inspection route, the image acquisition device 101 can be controlled to have its acquisition angle facing the battery array. In this embodiment, to reduce redundant image acquisition, multiple acquisition positions can be pre-set along the inspection route. The battery array images acquired by the image acquisition device 101 at multiple acquisition positions can cover the entire battery array of the data center. In this embodiment, multiple acquisition positions (corresponding to...) can be determined based on the acquisition angle of the image acquisition device 101 and the distance between the inspection route and the battery array. Figure 2 (The acquisition locations are determined in process 2). Alternatively, the acquisition range of the battery array images captured by the image acquisition device can be tested manually along the inspection route beforehand; and based on the manual test results, multiple acquisition locations can be determined so that the battery array images acquired by the image acquisition device 101 at multiple acquisition locations can cover the entire battery array of the data center. Each acquisition location is used to acquire images of a portion of the battery array in the data center.

[0046] Based on multiple pre-set data collection locations, the autonomous mobile device 10 can autonomously locate itself while moving along the battery array according to a set inspection route, thereby determining its current location. In this embodiment, the specific implementation method for autonomous positioning of the autonomous mobile device is not limited. In some embodiments, the autonomous mobile device 10 may employ SLAM technology for autonomous positioning.

[0047] Specifically, the autonomous mobile device 10 collects environmental information around its current location and determines its pose in the stored environmental map based on this information. Optionally, the autonomous mobile device 10 can construct a temporary map based on the acquired environmental information during movement and compare the constructed temporary map with the stored environmental map to determine the robot's pose in the stored environmental map. One possible implementation of comparing the constructed temporary map with the stored environmental map to determine the robot's pose in the stored environmental map is as follows: based on a matching algorithm, the constructed temporary map is traversed across each pose on the stored environmental map. For example, with a grid size of 5cm, a step size of 5cm can be selected. The temporary map covers all possible poses in the stored environmental map, and the angle step size is set to 5 degrees, including the orientation parameters of all poses. When a grid representing an obstacle on the temporary map matches a grid representing an obstacle on the stored environment map, a score is awarded, and the pose with the highest score is determined as the pose of the global optimal solution. Then, the matching rate of this global optimal solution pose is calculated, and if the matching rate of the global optimal solution pose is greater than a preset matching rate threshold, this global optimal solution pose is determined as the pose information of the autonomous mobile device 10. The pose information of the autonomous mobile device 10 includes: the position information and orientation information of the autonomous mobile device 10.

[0048] After the autonomous mobile device 10 determines its current location, it can match this location with multiple pre-defined acquisition locations to determine if the autonomous mobile device 10 has moved to a designated acquisition location. If the current location matches any of the pre-defined acquisition locations, the autonomous mobile device 10 is determined to have moved to the designated acquisition location P. Further, the autonomous mobile device 10 can acquire a battery array image A at the designated acquisition location P. This acquisition location P is the acquisition location where the autonomous mobile device 10 acquires the battery array image (corresponding to...). Figure 2 In process 2, the device moves to the acquisition location to acquire images of the battery array. The autonomous mobile device 10 can also record the correspondence between the battery array images and the acquisition location.

[0049] Considering the battery array has a certain height, multiple image acquisition devices 101 can be installed on the autonomous mobile device 10 to acquire images of the battery array at various heights. These multiple image acquisition devices 101 are installed at different heights on the autonomous mobile device 10, and each image acquisition device 101 at a different height is used to acquire images of the battery array corresponding to that height. Accordingly, when the autonomous mobile device 10 acquires images of the battery array at a set acquisition position P, it can obtain multiple images of the battery array located at different heights. Furthermore, the autonomous mobile device 10 can also record the correspondence between the battery array images, the acquisition position, and the height.

[0050] In other embodiments, the height of the image acquisition device 101 on the autonomous mobile device 10 is adjustable. Specifically, the autonomous mobile device 10 may include a retractable structure; the image acquisition device 101 is disposed on the retractable structure. The autonomous mobile device 10 adjusts the height of the image acquisition device 101 by adjusting the length of the retractable structure. In this embodiment, when the autonomous mobile device 10 moves to a set acquisition position P, it can adjust the height of the image acquisition device 101 at acquisition position P; and control the image acquisition device 101 to acquire battery array images at multiple heights, obtaining multiple battery array images at different heights. In the embodiments of this application, the battery array images acquired at the same height can be one or more. Multiple means two or more. Correspondingly, the autonomous mobile device 10 can also record the correspondence between the battery array images, acquisition positions, and heights.

[0051] In this embodiment, the autonomous mobile device 10 and the server device 20 can communicate with each other (corresponding to...). Figure 2 (Communication with the server device in process 1). The autonomous mobile device 10 and the server device 20 can be connected wirelessly or via a wired connection. Optionally, the autonomous mobile device 10 and the server device 20 can communicate via the Internet. Of course, the autonomous mobile device 10 and the server device 20 can also communicate with the terminal device 10b via a mobile network. Accordingly, the mobile network standard can be any one of 2G (GSM), 2.5G (GPRS), 3G (WCDMA, TD-SCDMA, CDMA2000, UTMS), 4G (LTE), 4G+ (LTE+), 5G, WiMax, etc. Optionally, the autonomous mobile device 10 and the server device 20 can also communicate via Bluetooth, WiFi, infrared, etc.

[0052] Based on the communication link between the autonomous mobile device 10 and the server device 20, the autonomous mobile device 10 can provide the acquired battery array image to the server device 20, which then performs anomaly detection and identification on the battery array based on the image. The server device 20 can be a single server device, a cloud-based server array, or a virtual machine (VM) running within a cloud-based server array. Alternatively, the server device 20 can also refer to other computing devices with corresponding service capabilities, such as computers or other terminal devices (running service programs). The process of the server device 20 performing anomaly detection and identification on the battery array based on the battery array image is illustrated below.

[0053] For storage batteries, battery abnormalities can be reflected to some extent on the battery's exterior. These abnormalities include, but are not limited to, leakage, bulging, corrosion, or acid creep. Therefore, in the embodiments of this application, as... Figure 1 As shown, the server device 20 can extract features from the received battery array image to determine the battery features of the battery array image. In this embodiment, the battery features of the battery array image refer to the external features of the battery reflected in the battery array image.

[0054] Since the battery features in the battery array image can reflect battery anomalies to a certain extent, the server device 20 can identify anomalies in the target batteries contained in the battery array image based on these features. This identifies the abnormal batteries and their anomaly types, achieving automatic identification of abnormal batteries and their anomaly types. Compared to manual inspection, this improves the efficiency of battery anomaly identification. Furthermore, compared to manual inspection, it avoids missed or misidentified battery anomalies due to insufficient human experience or negligence, reducing the probability of missed or misidentified anomalies. Here, "target battery" refers to the battery contained in the battery array image, and the number of target batteries can be one or more. "Multiple" means two or more. The specific number of target batteries is determined by the battery size, the acquisition angle of the image acquisition device, and the distance between the image acquisition device and the acquired image array. "Abnormal battery" refers to a battery in the target battery array that exhibits anomalies. The target battery array may or may not contain abnormal batteries. If abnormal batteries are present, they may be some or all of the batteries in the target battery array.

[0055] In the embodiments of this application, the specific implementation of extracting battery features from the battery array image and identifying anomalies in the target batteries contained in the battery array image is not limited.

[0056] In some embodiments, such as Figure 2 As shown, the server device 20 can perform battery detection on the battery array image to determine the local image corresponding to the target battery contained in the battery array image (corresponding to...). Figure 2 In process 3, battery detection is performed; then, battery features are extracted from local images to determine the battery features of the battery array image.

[0057] Optionally, the battery array image can be input into the battery detection model; in the battery detection model, battery detection is performed on the battery array image to obtain the spatial information of the target detection box (corresponding to) the battery array image with battery annotation. Figure 2 (Battery detection in process 2). In practical applications, when performing target detection on an image, rectangular detection boxes are typically used to label the target batteries contained in the image. The spatial information of the rectangular detection box specifically refers to the spatial information of the rectangular detection box on the battery array image, which reflects the spatial distribution of the target batteries in the battery array image. The spatial information of the rectangular detection box includes: the center position and size of the detection box. Optionally, the center position of the detection box can be represented by the center coordinates of the rectangular detection box, and the size of the detection box can be represented by the width and height of the rectangular detection box. Accordingly, the spatial information of the detection box can be represented as (x, y, w, h). Where (x, y) represents the center coordinates of the rectangular detection box, that is, the coordinates of the center of the rectangular detection box in the image to be detected, and w and h represent the width and height of the rectangular detection box, respectively. Alternatively, the spatial information of the rectangular detection box includes: the vertex coordinates of the rectangular detection box, etc. The center coordinates and vertex coordinates of the rectangular detection box are both coordinates on the battery array image.

[0058] A single frame of a battery array image can contain one or more target batteries. "Multiple" refers to two or more. Each target battery can correspond to one target detection box.

[0059] Furthermore, the server device 20 can extract the local image corresponding to the target detection box from the battery array image based on the spatial information of the target detection box; and input the local image into the battery anomaly recognition model; in the feature extraction layer of the battery anomaly recognition model, battery features can be extracted from the local image to obtain the battery features of the local image; further, the battery features of the local image can be input into the anomaly recognition layer of the battery anomaly recognition model. In the anomaly recognition layer, based on the battery features, the probability that the target battery contained in the local image belongs to multiple battery anomaly types can be calculated; based on the probability that the target battery belongs to multiple battery anomaly types, the abnormal battery in the target battery and the target anomaly type of the abnormal battery (corresponding to) can be determined. Figure 2(Battery anomaly identification in process 3). Optionally, for any target battery, if the probability of the target battery belonging to multiple battery anomaly types is less than a set anomaly probability threshold, then the target battery is determined to be a non-abnormal battery. If the probability of the target battery belonging to one or more battery anomaly types is greater than the set anomaly probability threshold, then the target battery is determined to be an abnormal battery. Further, for abnormal batteries, the battery anomaly type with the highest probability can be selected as the target anomaly type for the abnormal battery based on the probability of the abnormal battery belonging to multiple battery anomaly types.

[0060] In this embodiment, before using the battery detection model for battery detection and the battery anomaly identification model for battery anomaly type identification, both the battery detection model and the battery anomaly identification model need to be trained. The battery detection model and the battery anomaly identification model can be trained individually or jointly. Joint training of the battery detection model and the battery anomaly identification model refers to training both models using the same battery array sample images. The local image output by the battery detection model serves as the input to the battery anomaly identification model.

[0061] The battery detection model described above can be an Shot Detector (SSD) model, a YOLO (You Only Look Once) series model, a CenterNet model, an SPPNet model, an FPN model, etc., but is not limited to these. The battery anomaly identification model can be a neural network model, etc. The neural network model can be a CNN, RNN, or DNN model, etc.

[0062] In this embodiment, the battery array sample image can be a sample image of a battery with a known abnormal type. The battery array sample image set can be manually annotated to obtain the spatial information of the baseline detection box. In this embodiment, the baseline detection box refers to the detection box annotated with the batteries in the battery array sample image set.

[0063] Furthermore, such as Figure 2As shown in step 2, the battery detection model can be trained using a set of battery sample images to obtain the battery detection model. The initial model used for training the battery detection model is called the initial detection model. The initial detection model has the same model architecture as the final battery detection model obtained through model training, meaning the model parameters are the same. In this embodiment, model training mainly refers to training the parameters of the battery detection model using a set of battery array sample images to minimize the loss function. That is, with minimizing the loss function as the training objective, the model is trained using a set of battery array sample images to obtain the final battery detection model. The loss function can be determined based on the spatial information of the detection boxes obtained through model training and the spatial information of the baseline detection boxes from the battery array sample image set labeled with batteries before model training.

[0064] Optionally, the loss function L x It can be represented as:

[0065] L x =L center +λ scale L scale +λ offset L offset (3)

[0066] In the loss function (2), L center L represents the center loss, which is the loss between the center coordinates of the detection boxes obtained from model training and the center coordinates of the reference detection boxes; scale L represents the scale loss, which is the loss between the width and height of the detection box obtained from model training and the width and height of the baseline detection box; offset λ represents the offset loss, which is the offset of the center coordinates of the detection box obtained by the model training compared to the center coordinates of the reference detection box. scale , λ offset and λ θ These represent the weights of scale loss and offset loss, respectively, and can be flexibly set according to the actual situation.

[0067] After the battery detection model is trained, a local image set of the battery array sample image set containing the target battery, output by the trained battery detection model, can be used to train the battery anomaly recognition model. Specifically, minimizing the loss function can be the training objective, and the local image set can be used to train the battery anomaly recognition model to obtain the final trained battery anomaly recognition model. The loss function can be determined based on the difference between the predicted probability that the sample battery in the local image belongs to a preset number of anomaly types and the true probability of the sample battery in the local image belonging to the preset number of anomaly types. For example, the cross-entropy function between the predicted probability that the sample battery in the local image belongs to the preset number of anomaly types and the true probability of the sample battery in the local image belonging to the preset number of anomaly types can be used as the loss function. Since the anomaly types of the sample batteries in the local image are known, the true probability of the sample battery belonging to its determined anomaly type can be 1, and the true probability of belonging to other anomaly types can be 0. Accordingly, the loss function L... z It can be represented as:

[0068]

[0069] In equation (4), N represents the total number of battery array sample images. i represents the i-th frame of the battery array sample image; i = 1, 2, ..., N. M represents the total number of anomaly types; c represents the c-th anomaly type; c = 1, 2, ..., M. ic Let y represent the true probability that the sample batteries in the i-th frame of the battery array sample image belong to the c-th anomaly type. If the c-th anomaly type is a pre-labeled or determined anomaly type of the sample batteries contained in a local image of the i-th frame of the battery array sample image, then y ic =1; If the c-th anomaly type is not a pre-labeled or determined anomaly type of the sample battery contained in the local image of the i-th frame battery array sample image, then y ic =0. p ic This represents the predicted probability that a local image of the i-th frame of the battery array sample image output by the model training belongs to the c-th anomaly type.

[0070] The training process for the aforementioned battery detection model and battery anomaly recognition model can be executed on the server device 20 or on any other computing device. After the battery detection model and battery anomaly recognition model are trained, they can be pre-set in the server device 20. Thus, the server device 20 can use the battery detection model and battery anomaly recognition model to perform battery detection and battery anomaly recognition on the battery array image, obtaining the image coordinates of the target battery in the battery array image and the target anomaly type (corresponding to...) of the abnormal battery within the target battery. Figure 2 (Battery detection and battery anomaly identification in process 3).

[0071] In this embodiment, the server device 20 can determine the image coordinates of the abnormal battery in the battery array image based on the spatial information of the target detection box annotating the abnormal battery output by the battery detection model. For example, the center coordinates of the target detection box annotating the abnormal battery can be determined as the image coordinates of the abnormal battery in the battery array image. The center coordinates of the target detection box refer to the image coordinates of the center of the target detection box in the battery array image.

[0072] In this embodiment, to achieve battery inspection in the data center, it is also necessary to locate abnormal batteries, i.e., determine the location of the abnormal battery in the data center. This allows data center maintenance personnel to quickly find the abnormal battery for replacement or repair. To locate the abnormal battery in the data center, the server device 20 can also obtain the pose information of the autonomous mobile device during the battery array image acquisition process. Since the autonomous mobile device 10 can record the acquisition points, height, and orientation (i.e., posture) of the battery array image acquisition device when acquiring the battery array image, and when providing the battery array image to the server device 20, the autonomous mobile device 10 can also provide the acquisition points, height, and orientation of the image acquisition device corresponding to the battery array image to the server device 20. Therefore, the server device 20 can receive the battery array image, as well as the acquisition points, height, and orientation of the image acquisition device corresponding to the battery array image; that is, the server device 20 can obtain the acquisition points and the height and orientation of the image acquisition device recorded by the autonomous mobile device 10 when acquiring the battery array image.

[0073] Furthermore, the server device 20 can determine the pose information of the autonomous mobile device 10 during the process of acquiring battery array images based on the acquisition points recorded by the autonomous mobile device 10 when acquiring battery array images, as well as the height and posture of the image acquisition device.

[0074] Furthermore, the server device 20 can determine the position information of the abnormal battery in the set coordinate system (corresponding to the position information of the abnormal battery in the battery array image) based on the pose information of the autonomous mobile device 10 during the acquisition of battery array images and the image coordinates of the abnormal battery in the battery array image. Figure 2 In process 4, the location of the abnormal battery in the anomaly identification result is determined based on the collection location, height, and center coordinates. In this embodiment, the reference object for establishing the coordinate system is not limited. In some embodiments, the coordinate system may be a world coordinate system; or, a coordinate system established with the data center as the reference object; or, a coordinate system established with the battery room in the data center as the reference object, etc. The coordinate system corresponds to the data center, and each coordinate in the coordinate system corresponds to a certain location (three-dimensional spatial location) in the data center. Therefore, the location information of the abnormal battery in the data center can be determined based on the location information of the abnormal battery under the set coordinates, thus realizing the localization of the abnormal battery in the data center.

[0075] In this embodiment, to ensure that data center operations and maintenance personnel can perceive the reliability of the inspection results, the server device 20 can also determine the confidence level that the abnormal battery's anomaly type is the target anomaly type based on the probability that the abnormal battery belongs to the target anomaly type determined by the battery anomaly identification model. Specifically, the probability that the abnormal battery belongs to the target anomaly type can be used as the confidence level that the abnormal battery's anomaly type is the target anomaly type.

[0076] Furthermore, the server-side device 20 can identify the anomaly of the target battery based on the target anomaly type, the confidence level that the anomaly battery belongs to the target anomaly type, and the location information of the anomaly battery in the data center. This anomaly identification result includes: the target anomaly type of the anomaly battery, the confidence level that the anomaly battery belongs to the target anomaly type, and the location information of the anomaly battery in the data center.

[0077] In this embodiment of the application, in order to improve the accuracy of the anomaly identification results, the anomaly identification results can be screened to filter out anomaly identification results that do not meet the requirements. Specifically, the server device 20 can screen the anomaly identification results of the abnormal battery based on the confidence level that the abnormal battery is the target anomaly type; if the confidence level that the abnormal battery is the target anomaly type is less than the set confidence level threshold, it is determined that the anomaly identification result of the abnormal battery does not meet the set confidence level requirement, and the anomaly identification result can be filtered out.

[0078] Optionally, the server device 20 can also filter the anomaly identification results of abnormal batteries based on the spatial information of the target detection box annotated with abnormal batteries determined by the battery detection model. This is mainly because, in order to improve the accuracy of the anomaly identification results, when the autonomous mobile device 10 acquires images of the battery array, the battery array located at the center of the acquisition view of the image acquisition device 101 is taken as the battery array to be detected. This ensures that the battery array to be detected is located in the central area of ​​the acquired battery array image, while the battery arrays located in the edge areas of the battery array image are not the focus of this acquisition, and only partial images of these battery arrays may be acquired. Therefore, the accuracy of the anomaly identification results for these battery arrays in the edge areas of the battery array image is low. Based on this, a target image coordinate range can be preset, which is used to filter battery arrays located in the more central area of ​​the battery array image. Based on the preset target image coordinate range, it can be determined whether the center coordinates of the spatial information of the target detection box annotated with abnormal batteries determined by the battery detection model are within the preset target image coordinate range; if the determination result is yes, the anomaly identification result of the abnormal battery is retained; if the determination result is no, the anomaly identification result of the abnormal battery is filtered out.

[0079] The above-described methods of filtering anomaly identification results based on the confidence level of the abnormal battery as the target anomaly type, and filtering anomaly identification results based on the spatial information of the target detection box annotated with the abnormal battery determined by the battery detection model, can be implemented individually or in combination. When implemented in combination, the anomaly identification results of the target battery can be filtered based on the spatial information of the target detection box annotated with the abnormal battery determined by the battery detection model and the confidence level of the abnormal battery as the target anomaly type. Specifically, it can be determined whether the confidence level of the target battery as the target anomaly type is greater than or equal to a set confidence threshold; and whether the center coordinates in the spatial information of the target detection box determined by the target detection model are within the set target image coordinate range; if both determinations are yes, the anomaly identification results of the target battery are retained; if any determination result is no, the anomaly identification results of the target battery are filtered to obtain the anomaly identification results of the target battery that meet the requirements (corresponding to...). Figure 2 In process 3, anomaly detection results are filtered based on the center coordinates and confidence level of the target detection box.

[0080] Furthermore, the server device 20 can output the anomaly identification result of the abnormal battery that meets the requirements. In this embodiment, the specific implementation form of the anomaly identification result output by the server device 20 is not limited. In some embodiments, the server device 20 can send the anomaly identification result of the abnormal battery to the computing device (such as a terminal device) of the maintenance personnel. The maintenance personnel's computing device receives the anomaly identification result of the abnormal battery and can display the anomaly identification result of the abnormal battery on the screen. In other embodiments, the server device 20 can also send the anomaly identification result of the abnormal battery to a data center management device. The data center management device receives the anomaly identification result of the abnormal battery and can display the anomaly identification result of the target battery on the screen, etc.

[0081] For each frame of battery array image collected by the server-side device 20 during the inspection of the data center by the autonomous mobile device, anomaly identification can be performed on the target batteries included, obtaining the anomaly identification results of the target batteries included in each frame of the battery array image; and based on the anomaly identification results of the target batteries included in each frame of the battery array image, a battery inspection anomaly detection report of the data center can be generated (corresponding to...). Figure 2 Battery inspection report in process 4: location of abnormal battery, type of abnormality, and confidence level.

[0082] Optionally, when generating the battery inspection anomaly detection report for the data center, the server device 20 can also perform deduplication processing on the anomaly identification results of the target batteries contained in each frame of the battery array image. For example, the server device 20 can obtain the location information of the target batteries in each frame of the battery array image within the data center based on the anomaly identification results of the target batteries contained in each frame of the battery array image, and remove duplicate location information of the target batteries. Then, based on the deduplicated anomaly identification results, the battery inspection anomaly detection report for the data center is generated. The battery inspection anomaly detection report for the data center can be displayed in chart form. The information items contained in a row or column of the chart may include: the location information of the target battery in the data center, the anomaly type of the target battery, and the confidence level that the target battery belongs to that anomaly type, etc.

[0083] Furthermore, server-side device 20 can output a battery inspection anomaly detection report for the data center (corresponding to...). Figure 2 In process 4, the data center management system outputs a battery inspection report. For details on the specific implementation of the server-side device 20 outputting the data center's battery inspection anomaly detection report, please refer to the relevant content regarding the anomaly identification results of the target battery output by the server-side device 20, which will not be repeated here. Data center battery maintenance personnel can query the data center's battery inspection anomaly detection report to determine information about abnormal batteries, etc.

[0084] It is worth noting that the above-mentioned implementation method of server device 20 performing battery anomaly identification based on battery array images can also be deployed on autonomous mobile device 10, whereby autonomous mobile device 10 autonomously completes the processes of acquiring battery array images and autonomously identifying battery anomalies based on the battery array images. This process does not require the participation of server device 20. For the specific implementation method of autonomous mobile device 10 performing battery anomaly identification based on battery array images, please refer to the relevant content of server device 20 performing battery anomaly identification based on battery array images, which will not be repeated here.

[0085] In addition to the data center inspection system provided in the above embodiments, this application also provides a battery anomaly identification method. The battery anomaly identification method provided in this application will be described exemplarily below with reference to specific embodiments.

[0086] Figure 3 This is a flowchart illustrating the battery anomaly identification method provided in an embodiment of this application. Figure 3 As shown, the method mainly includes:

[0087] 301. Acquire images of the battery array as the autonomous mobile device moves along the battery array in the data center.

[0088] 302. Extract battery features from the battery array image to determine the battery features of the battery array image.

[0089] 303. Based on battery characteristics, perform anomaly identification on the target batteries contained in the battery array image to determine the abnormal batteries in the target batteries and the target anomaly type of the abnormal batteries.

[0090] The battery anomaly identification method provided in this embodiment can be executed by any device with computing capabilities, including but not limited to: autonomous mobile devices and server devices.

[0091] In step 301 of this embodiment, battery array images can be acquired as the autonomous mobile device moves along the battery array in the data center. For a description of the autonomous mobile device acquiring battery array images, please refer to the relevant content in the above system embodiment, which will not be repeated here. For batteries, battery abnormalities can be reflected to some extent on the battery's exterior. These abnormality types include, but are not limited to, leakage, bulging, corrosion, or acid creep. Therefore, in step 302, feature extraction can be performed on the battery array images to determine the battery characteristics of the battery array images. In this embodiment, the battery characteristics of the battery array images refer to the external features of the battery reflected in the battery array images.

[0092] Since the battery features in the battery array image can reflect battery anomalies to a certain extent, in step 303, anomaly identification can be performed on the target batteries contained in the battery array image based on the battery features reflected in the image. This identifies the abnormal batteries and their target anomaly types, achieving automatic identification of abnormal batteries and their anomaly types. Compared to manual inspection, this helps improve the efficiency of battery anomaly identification. On the other hand, compared to manual inspection, it avoids missed or misidentified battery anomalies due to insufficient human experience or human negligence, helping to reduce the probability of missed or misidentified battery anomalies.

[0093] In the embodiments of this application, the specific implementation of extracting battery features from the battery array image and identifying anomalies in the target batteries contained in the battery array image is not limited.

[0094] In some embodiments, battery detection can be performed on the battery array image to determine a local image corresponding to a target battery contained in the battery array image; then, battery feature extraction can be performed on the local image to determine the battery features of the battery array image.

[0095] Optionally, the battery array image can be input into a battery detection model; in the battery detection model, battery detection is performed on the battery array image to obtain spatial information of the target detection boxes for battery annotation in the battery array image. For a description of the spatial information of the target detection boxes, please refer to the relevant content of the above system embodiment, which will not be repeated here. A single frame of the battery array image can contain one or more target batteries. "Multiple" refers to two or more. Each target battery can correspond to one target detection box.

[0096] Furthermore, based on the spatial information of the target detection box, a local image corresponding to the target detection box can be extracted from the battery array image; this local image is then input into the battery anomaly recognition model; in the feature extraction layer of the battery anomaly recognition model, battery features can be extracted from the local image to obtain the battery features of the local image; further, the battery features of the local image can be input into the anomaly recognition layer of the battery anomaly recognition model. In the anomaly recognition layer, based on the battery features, the probability that the target battery contained in the local image belongs to multiple battery anomaly types can be calculated; based on the probability that the target battery belongs to multiple battery anomaly types, the abnormal battery in the target battery and the target anomaly type of the abnormal battery are determined. Optionally, for any target battery, if the probability that the target battery belongs to multiple battery anomaly types is less than a set anomaly probability threshold, then the target battery is determined to be a non-abnormal battery. If the probability that the target battery belongs to one or more battery anomaly types is greater than the set anomaly probability threshold, then the target battery is determined to be an abnormal battery. Further, for an abnormal battery, based on the probability that the abnormal battery belongs to multiple battery anomaly types, the battery anomaly type with the highest probability can be selected as the target anomaly type of the abnormal battery.

[0097] In this embodiment, before using the battery detection model for battery detection and the battery anomaly identification model for battery anomaly type identification, it is necessary to train both the battery detection model and the battery anomaly identification model. The training process for the battery detection model and the battery anomaly identification model can be found in the relevant content of the above system embodiment, and will not be repeated here.

[0098] After the battery detection model and battery anomaly recognition model are trained, they can be pre-set on the computing device. This allows the battery detection model and battery anomaly recognition model to be used to perform battery detection and battery anomaly recognition on the battery array image, obtaining the image coordinates of abnormal batteries within the target battery array image and the target anomaly type of the abnormal battery.

[0099] In this embodiment, the image coordinates of the abnormal battery in the battery array image can be determined based on the spatial information of the target detection box annotating the abnormal battery output by the battery detection model. For example, the center coordinates of the target detection box annotating the abnormal battery can be determined as the image coordinates of the abnormal battery in the battery array image. The center coordinates of the target detection box refer to the image coordinates of the center of the target detection box in the battery array image.

[0100] In this embodiment, to achieve battery inspection in the data center, it is also necessary to locate abnormal batteries, i.e., determine their location within the data center. This allows maintenance personnel to quickly locate the abnormal batteries for replacement or repair. To locate the abnormal batteries within the data center, the pose information of the autonomous mobile device during the acquisition of battery array images can also be obtained. Since the autonomous mobile device can record the acquisition points, height, and orientation (i.e., posture) of the battery array images during acquisition, the acquisition points, height, and posture of the image acquisition device recorded by the autonomous mobile device during battery array image acquisition can be obtained.

[0101] Furthermore, the pose information of the autonomous mobile device during the battery array image acquisition process can be determined based on the acquisition points recorded by the autonomous mobile device when acquiring battery array images, as well as the height and orientation of the image acquisition device.

[0102] Furthermore, based on the pose information of the autonomous mobile device during the acquisition of battery array images, and the image coordinates of the abnormal battery in the battery array image, the position information of the battery marked as abnormal can be determined in the set coordinate system; and based on the position information of the abnormal battery in the set coordinate system, the position information of the abnormal battery in the data center can be determined, thereby realizing the location of the abnormal battery in the data center.

[0103] In this embodiment of the application, to enable data center operations and maintenance personnel to perceive the reliability of inspection results, the confidence level of the abnormal battery's anomaly type as the target anomaly type can be determined based on the probability that the abnormal battery belongs to the target anomaly type as determined by the battery anomaly identification model. Specifically, the probability that the abnormal battery belongs to the target anomaly type can be used as the confidence level of the abnormal battery's anomaly type as the target anomaly type.

[0104] Furthermore, the anomaly identification result of the abnormal battery can be determined based on the target anomaly type of the abnormal battery, the confidence level that the abnormal battery belongs to the target anomaly type, and the location information of the abnormal battery in the data center. This anomaly identification result includes: the target anomaly type of the abnormal battery, the confidence level that the abnormal battery belongs to the target anomaly type, and the location information of the abnormal battery in the data center.

[0105] In this embodiment of the application, in order to improve the accuracy of the anomaly identification results, the anomaly identification results can be screened to filter out anomaly identification results that do not meet the requirements. Specifically, the anomaly identification results of the target battery can be screened according to the confidence level that the abnormal battery is the target anomaly type; if the confidence level that the abnormal battery is the target anomaly type is less than the set confidence level threshold, it is determined that the anomaly detection result of the abnormal battery does not meet the set confidence level requirement, and the anomaly identification result can be filtered out.

[0106] Optionally, the spatial information of the target detection boxes for annotating abnormal batteries can be determined based on the battery detection model, and the anomaly identification results of abnormal batteries can be filtered. This is mainly because, in order to improve the accuracy of the anomaly identification results, a target image coordinate range can be preset. This target image coordinate range is used to filter battery arrays located in the relatively central area of ​​the battery array image. Based on the preset target image coordinate range, it can be determined whether the center coordinates of the spatial information of the target detection boxes for annotating abnormal batteries determined by the target detection model are within the preset target image coordinate range; if the determination result is yes, the anomaly identification result of the abnormal battery is retained; if the determination result is no, the anomaly identification result of the abnormal battery is filtered out.

[0107] The above-described implementation methods—filtering anomaly identification results based on the confidence level of the abnormal battery as the target anomaly type, and filtering anomaly identification results based on the spatial information of the target detection box annotated with the abnormal battery determined by the battery detection model—can be implemented individually or in combination. When implemented in combination, the anomaly identification results of the abnormal battery can be filtered based on the spatial information of the target detection box annotated with the abnormal battery determined by the battery detection model and the confidence level of the abnormal battery as the target anomaly type. Specifically, it can be determined whether the confidence level of the abnormal battery as the target anomaly type is greater than or equal to a set confidence threshold; and whether the center coordinates in the spatial information of the target detection box annotated with the abnormal battery determined by the target detection model are within a set target image coordinate range. If both determinations are yes, the anomaly identification results of the abnormal battery are retained; if any determination result is no, the anomaly identification results of the abnormal battery are filtered to obtain the anomaly identification results of the abnormal battery that meet the requirements.

[0108] Furthermore, the abnormal battery identification result that meets the requirements can be output. In this application embodiment, the specific implementation form of outputting the abnormal battery identification result is not limited. In some embodiments, the execution subject of the above-described battery abnormality identification method is a server-side device, which can send the abnormal battery identification result to the operation and maintenance personnel's computing device (such as a terminal device). The operation and maintenance personnel's computing device receives the abnormal battery identification result and can display the abnormal battery identification result on the screen. In other embodiments, the abnormal battery identification result can also be sent to a data center management device. The data center management device receives the abnormal battery identification result and can display the abnormal battery identification result on the screen, etc. In still other embodiments, the execution subject of the above-described battery abnormality identification method is an autonomous mobile device. Accordingly, the autonomous mobile device can display the abnormal identification result of the target battery.

[0109] In this embodiment of the application, for each frame of battery array image collected by the autonomous mobile device during the inspection of the data center, anomaly identification can be performed on the target battery contained therein, and the anomaly identification result of the target battery contained in each frame of battery array image can be obtained; and a battery inspection anomaly detection report of the data center can be generated based on the anomaly identification result of the target battery contained in each frame of battery array image.

[0110] Optionally, when generating a battery inspection anomaly detection report for the data center, the anomaly identification results can be deduplicated based on the anomaly identification results of the abnormal batteries in the target batteries contained in each frame of the battery array image. For example, based on the anomaly identification results of the abnormal batteries in the target batteries contained in each frame of the battery array image, the location information of the abnormal batteries in the target batteries contained in each frame of the battery array image within the data center can be obtained, and duplicate location information of abnormal batteries can be removed. Then, based on the deduplicated anomaly identification results, a battery inspection anomaly detection report for the data center is generated. The battery inspection anomaly detection report for the data center can be displayed in chart form. The information items contained in a row or column of the chart may include: the location information of the abnormal battery in the data center, the anomaly type of the abnormal battery, and the confidence level that the abnormal battery belongs to that anomaly type, etc.

[0111] Furthermore, it can output battery inspection anomaly detection reports for the data center. For specific implementation details on outputting these reports, please refer to the section on outputting the anomaly identification results of the target battery, which will not be repeated here. Data center battery maintenance personnel can query these reports to identify information about abnormal batteries, etc.

[0112] It should be noted that the execution subject of each step of the method provided in the above embodiments can be the same device, or the method can be executed by different devices. For example, the execution subject of steps 301 and 302 can be device A; or the execution subject of step 301 can be device A, and the execution subject of step 302 can be device B; and so on.

[0113] Furthermore, some processes described in the above embodiments and accompanying drawings include multiple operations that appear in a specific order. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or they may be executed in parallel. The operation numbers, such as 301, 302, etc., are merely used to distinguish different operations and do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel.

[0114] Accordingly, embodiments of this application also provide a computer-readable storage medium storing computer instructions, which, when executed by one or more processors, cause one or more processors to perform the steps in the above-described battery anomaly identification method.

[0115] This application also provides a computer program product, including a computer program. The computer program product is executed by a processor to implement the above-described battery anomaly identification method. In this application, the specific implementation form of the computer program product is not limited. In some embodiments, the computer program product may be implemented as a plug-in or other functional module, or as a SaaS software product deployed in the cloud to provide battery anomaly identification services, etc.

[0116] Figure 4 This is a schematic diagram of the structure of a computing device provided in an embodiment of this application. Figure 4 As shown, the computing device includes a memory 40a, a processor 40b, and a communication component 40c. The memory 40a is used to store computer programs.

[0117] The processor 40b is coupled to the memory 40a for executing a computer program for: acquiring battery array images collected by the autonomous mobile device as it moves along the battery array in the data center via the communication component 40c; extracting battery features from the battery array images to determine the battery features of the battery array images; and, based on the battery features, identifying anomalies in the target batteries contained in the battery array images to determine the abnormal batteries in the target batteries and the target anomaly type of the abnormal batteries.

[0118] When performing battery identification on a battery array image, the processor 40b specifically performs the following: inputs the battery array image into a battery detection model to obtain spatial information of the target detection box that annotates the battery in the battery array image; extracts the local image corresponding to the target detection box from the battery array image based on the spatial information of the target detection box; and extracts battery features from the local image to determine the battery features of the battery array image.

[0119] Optionally, when extracting battery features from a local image, the processor 40b is specifically used to: input the local image into the feature extraction layer of the battery anomaly recognition model to obtain the battery features of the local image.

[0120] Furthermore, when the processor 40b performs anomaly identification on the target battery contained in the battery array image based on the battery features, it specifically performs the following: inputting the battery features into the anomaly identification layer in the battery anomaly identification model; in the anomaly identification layer, calculating the probability that the target battery contained in the local image belongs to multiple battery anomaly types based on the battery features; and determining the abnormal battery in the target battery and the target anomaly type of the abnormal battery based on the probability that the target battery belongs to multiple battery anomaly types.

[0121] In this embodiment, the processor 40b is further configured to: acquire pose information of the autonomous mobile device during the acquisition of battery array images; determine the image coordinates of the abnormal battery in the battery array image based on the spatial information of the target detection box annotated with the abnormal battery; determine the position information of the abnormal battery in a set coordinate system based on the pose information of the autonomous mobile device during the acquisition of battery array images and the image coordinates of the abnormal battery in the battery array image; and determine the position information of the abnormal battery in the data center based on the position information of the abnormal battery in the set coordinate system.

[0122] Optionally, when the processor 40b acquires the pose information of the autonomous mobile device during the process of acquiring battery array images, it is specifically used to: acquire the acquisition points recorded by the autonomous mobile device during the acquisition of battery array images, as well as the height and orientation of the image acquisition device; and determine the pose information of the autonomous mobile device during the process of acquiring battery array images based on the acquisition points recorded by the autonomous mobile device during the acquisition of battery array images, as well as the height and orientation of the image acquisition device.

[0123] Optionally, the processor 40b is further configured to: determine the confidence level of the abnormal battery as the target abnormal type based on the probability that the abnormal battery belongs to the target abnormal type; determine the abnormal identification result of the abnormal battery based on the target abnormal type of the abnormal battery, the confidence level of the abnormal battery as the target abnormal type, and the location information of the abnormal battery in the data center; and output the abnormal identification result of the abnormal battery.

[0124] Optionally, the processor 40b is further configured to: determine the confidence level of a target battery as a target anomaly type based on the probability that the anomaly battery belongs to the target anomaly type; and filter the anomaly identification results of the anomaly battery based on the spatial information of the target detection box marked on the anomaly battery and the confidence level of the anomaly battery as a target anomaly type, so as to filter out the anomaly identification results that do not meet the requirements. Optionally, the processor 40b is further configured to: generate a battery inspection report for the data center based on the anomaly identification results of the target batteries contained in each frame of battery array image collected by the autonomous mobile device during the inspection of the data center; and output the battery inspection report.

[0125] In some alternative implementations, such as Figure 4 As shown, the computing device may also include optional components such as a power supply component 40d. Figure 4 The diagram only shows some components and does not mean that the computing device must contain them. Figure 4 The inclusion of all components does not imply that a computing device can only include... Figure 4 The components shown.

[0126] The computing device provided in this embodiment can acquire battery array images of a data center collected by an autonomous mobile device; extract battery features from the battery array images to obtain the battery characteristics reflected in the images; then, based on the battery characteristics reflected in the images, anomaly identification can be performed on the target batteries contained in the images to determine the abnormal batteries and their target anomaly types. This achieves automatic identification of abnormal batteries and their anomaly types, which helps improve the efficiency of battery anomaly identification compared to manual inspection. Furthermore, compared to manual inspection, it avoids missed or misidentified battery anomalies due to insufficient human experience or negligence, helping to reduce the probability of missed or misidentified battery anomalies.

[0127] Figure 5 This is a schematic diagram of the structure of an autonomous mobile device provided in an embodiment of this application. Figure 5 As shown, the autonomous mobile device includes: a mechanical body 501; an image acquisition device 502, a memory 503 and a processor 504 are installed on the mechanical body 501; the memory 503 is used to store computer programs.

[0128] It is worth noting that the number of memory 503 and processor 504 can be one or more. "More than" refers to two or more. In this embodiment, memory 503 and processor 504 can be disposed inside the mechanical body 501 or on the surface of the mechanical body 501. Image acquisition device 502 is disposed on the surface of the mechanical body.

[0129] The mechanical body 501 is the actuator of the autonomous mobile device, capable of executing operations specified by the processor 102 within a defined environment. The mechanical body 501, to a certain extent, reflects the physical form of the autonomous mobile device. However, in this embodiment, the physical form of the autonomous mobile device is not limited. The mechanical body 501 primarily refers to the body of the autonomous mobile device.

[0130] It is worth noting that the mechanical body 501 also includes some basic components of the autonomous mobile device, such as a drive component, a communication component, a power supply component, a display component, an audio component, etc. Optionally, the drive component may include drive wheels, drive motors, casters, etc. The basic components and their configurations may vary in different autonomous mobile devices; the embodiments listed in this application are only some examples.

[0131] In this embodiment, the image acquisition device 502 can acquire images of the battery array as the autonomous mobile device moves along the battery array in the data center.

[0132] In some embodiments, the memory 503 stores a set inspection route and a set acquisition location. The processor 504 can be used to: control an autonomous mobile device to move along the battery array in the data center according to the set inspection route; and when it moves to the set acquisition location, control the image acquisition device 502 to acquire images of the battery array at the acquisition location.

[0133] Optionally, the processor 504 is further configured to: autonomously locate the autonomous mobile device during its movement to determine its location information; and match the location information of the autonomous mobile device with a set collection point; if a match is found, determine that the autonomous mobile device has moved to a set collection location.

[0134] In other embodiments, a retractable structure 505 is provided on the autonomous mobile device. An image acquisition device 502 is disposed on the retractable structure 505. When the processor 504 controls the image acquisition device 502 to acquire battery array images at the acquisition position, it specifically performs the following: adjusts the length of the retractable structure 505 at the acquisition position to adjust the height of the image acquisition device 502; and controls the image acquisition device 502 to acquire battery array images at multiple heights.

[0135] In this embodiment, the processor 504 is coupled to the memory 503 to execute a computer program for: extracting battery features from a battery array image to determine the battery features of the battery array image; and, based on the battery features, identifying anomalies in the target batteries contained in the battery array image to determine the abnormal batteries in the target batteries and the target anomaly type of the abnormal batteries.

[0136] When performing battery identification on a battery array image, the processor 504 specifically performs the following steps: inputs the battery array image into a battery detection model to obtain spatial information of the target detection box that annotates the battery in the battery array image; extracts the local image corresponding to the target detection box from the battery array image based on the spatial information of the target detection box; and extracts battery features from the local image to determine the battery features of the battery array image.

[0137] Optionally, when extracting battery features from a local image, the processor 504 specifically performs the following: inputting the local image into the feature extraction layer of the battery anomaly recognition model to obtain the battery features of the local image.

[0138] Furthermore, when the processor 504 performs anomaly identification on the target battery contained in the battery array image based on the battery features, it specifically performs the following: inputting the battery features into the anomaly identification layer in the battery anomaly identification model; in the anomaly identification layer, calculating the probability that the target battery contained in the local image belongs to multiple battery anomaly types based on the battery features; and determining the abnormal battery in the target battery and the target anomaly type of the abnormal battery based on the probability that the target battery belongs to multiple battery anomaly types.

[0139] In this embodiment, the processor 504 is further configured to: acquire pose information of the autonomous mobile device during the acquisition of battery array images; determine the image coordinates of the abnormal battery in the battery array image based on the spatial information of the target detection box annotated with the abnormal battery; determine the position information of the abnormal battery in a set coordinate system based on the pose information of the autonomous mobile device during the acquisition of battery array images and the image coordinates of the abnormal battery in the battery array image; and determine the position information of the abnormal battery in the data center based on the position information of the abnormal battery in the set coordinate system.

[0140] Optionally, when the processor 504 acquires the pose information of the autonomous mobile device during the process of acquiring battery array images, it is specifically used to: acquire the acquisition points recorded by the autonomous mobile device during the acquisition of battery array images, as well as the height and orientation of the image acquisition device; and determine the pose information of the autonomous mobile device during the process of acquiring battery array images based on the acquisition points recorded by the autonomous mobile device during the acquisition of battery array images, as well as the height and orientation of the image acquisition device.

[0141] Optionally, the processor 504 is further configured to: determine the anomaly identification result of the abnormal battery based on the probability that the abnormal battery belongs to the target anomaly type and the confidence level that the abnormal target battery is the target anomaly type; determine the anomaly identification result of the abnormal battery based on the target anomaly type of the abnormal battery, the confidence level that the target battery is the target anomaly type and the location information of the abnormal battery in the data center; and output the anomaly identification result of the abnormal battery.

[0142] Optionally, the processor 504 is further configured to: determine the confidence level of an abnormal battery as a target abnormal type based on the probability that the abnormal battery belongs to the target abnormal type; and filter the abnormal battery identification results based on the spatial information of the target detection box annotated with the abnormal battery and the confidence level of the abnormal battery as a target abnormal type, so as to filter out abnormal identification results that do not meet the requirements. Optionally, the processor 504 is further configured to: generate a battery inspection report for the data center based on the abnormal identification results of the target batteries contained in each frame of battery array image collected by the autonomous mobile device during the inspection of the data center; and output the battery inspection report.

[0143] It should be noted that, Figure 5 The diagram only shows some components and does not mean that autonomous mobile devices must include them. Figure 5 The inclusion of all components does not imply that autonomous mobile devices can only include... Figure 5 The components shown.

[0144] The autonomous mobile device provided in this embodiment can acquire battery array images of a data center collected by the autonomous mobile device; extract battery features from the battery array images to obtain the battery features reflected in the images; then, based on the battery features reflected in the images, anomaly identification can be performed on the target batteries contained in the images to determine the target anomaly type of the target battery. This achieves automatic identification of abnormal batteries and their anomaly types, which helps improve the efficiency of battery anomaly identification compared to manual inspection. On the other hand, compared to manual inspection, it avoids missed or misidentified battery anomalies due to insufficient human experience or human negligence, helping to reduce the probability of missed or misidentified battery anomalies.

[0145] In this embodiment, the memory is used to store computer programs and can be configured to store various other data to support operation on its host device. The processor can execute the computer programs stored in the memory to implement corresponding control logic. The memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0146] In the embodiments of this application, the processor can be any hardware processing device capable of executing the above-described method logic. Optionally, the processor can be a central processing unit (CPU), a graphics processing unit (GPU), or a microcontroller unit (MCU); it can also be a field-programmable gate array (FPGA), a programmable array logic (PAL), a general array logic (GAL), a complex programmable logic device (CPLD), or other programmable devices; or it can be an advanced reduced instruction set (RISC) processor (ARM) or a system on chip (SOC), etc., but is not limited thereto.

[0147] In this embodiment, the communication component is configured to facilitate wired or wireless communication between its host device and other devices. The device housing the communication component can access wireless networks based on communication standards, such as WiFi, 2G or 3G, 4G, 5G, or combinations thereof. In one exemplary embodiment, the communication component receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In another exemplary embodiment, the communication component may also be implemented based on Near Field Communication (NFC), Radio Frequency Identification (RFID), Infrared Data Association (IrDA), Ultra Wideband (UWB), Bluetooth (BT), or other technologies.

[0148] In embodiments of this application, the display component may include a liquid crystal display (LCD) and a touch panel (TP). If the display component includes a touch panel, the display component may be implemented as a touchscreen to receive input signals from a user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors may not only sense the boundaries of touch or swipe actions but also detect the duration and pressure associated with the touch or swipe operation.

[0149] In this embodiment, a power supply component is configured to provide power to various components of the device in which it resides. The power supply component may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device in which the power supply component resides.

[0150] In embodiments of this application, the audio component can be configured to output and / or input audio signals. For example, the audio component includes a microphone (MIC), which is configured to receive external audio signals when the device containing the audio component is in an operating mode, such as call mode, recording mode, and voice recognition mode. The received audio signals can be further stored in memory or transmitted via a communication component. In some embodiments, the audio component also includes a speaker for outputting audio signals. For example, in devices with voice interaction capabilities, voice interaction with the user can be achieved through the audio component.

[0151] It should be noted that the terms "first" and "second" in this article are used to distinguish different messages, devices, modules, etc., and do not represent a chronological order, nor do they limit "first" and "second" to different types.

[0152] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0153] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0154] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0155] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0156] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0157] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0158] Computer storage media are readable storage media, also known as removable media. Removable and non-removable media can be used to store information by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media do not include transient media, such as modulated data signals and carrier waves.

[0159] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0160] The above description is merely an embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of this application should be included within the scope of the claims of this application.

Claims

1. A data center inspection system, comprising: include: Independent mobile devices and server-side devices; The autonomous mobile device is equipped with an image acquisition device; The autonomous mobile device is used to move along the battery array in the data center, and during the movement along the battery array, controls the image acquisition device to acquire images of the battery array; and provides the battery array images to the server device. The server-side device is configured to: extract battery features from the battery array image to determine the battery features of the battery array image; and, based on the battery features, identify anomalies in the target batteries contained in the battery array image to determine the abnormal batteries in the target batteries and the target anomaly type of the abnormal batteries. And, obtain the pose information of the autonomous mobile device during the process of acquiring the battery array image; The pose information includes: position information and orientation information; based on the spatial information of the target detection box annotating the abnormal battery, the image coordinates of the abnormal battery in the battery array image are determined; the spatial information includes: the center coordinates of the target detection box; based on the pose information of the autonomous mobile device during the acquisition of the battery array image and the image coordinates of the abnormal battery in the battery array image, the position information of the abnormal battery in a set coordinate system is determined; and based on the position information of the abnormal battery in the set coordinate system, the position information of the abnormal battery in the data center is determined; and if the center coordinates are within the set target image coordinate range, and the confidence that the abnormal battery is the target abnormal type is greater than or equal to the set confidence threshold, then the abnormal identification result of the abnormal battery is retained; or, if the center coordinates are not within the set target image coordinate range, and / or, the confidence that the abnormal battery is the target abnormal type is less than the set confidence threshold, then the abnormal identification result of the abnormal battery is filtered. The target image coordinate range is used to filter battery arrays located in the central region of the battery array image.

2. The system of claim 1, wherein, When the autonomous mobile device moves along the battery array in the data center, it is specifically used for: The system moves along the battery array within the data center according to the pre-defined inspection route. When the autonomous mobile device controls the image acquisition device to acquire images of the battery array, it is specifically used for: When the device moves to the set acquisition position, it controls the image acquisition device to acquire images of the battery array at that acquisition position.

3. The system of claim 2, wherein, When the autonomous mobile device controls the image acquisition device to acquire images of the battery array at the acquisition location, it is specifically used for: Adjust the height of the image acquisition device at the acquisition location; The image acquisition device is controlled to acquire images of the battery array at multiple heights.

4. A battery abnormality recognition method characterized by comprising: include: Acquire images of the battery array as an autonomous mobile device moves along the battery array in a data center; Battery features are extracted from the battery array image to determine the battery features of the battery array image; Based on the battery characteristics, anomaly identification is performed on the target batteries contained in the battery array image to determine the abnormal batteries in the target batteries and the target anomaly type of the abnormal batteries; as well as, Acquire the pose information of the autonomous mobile device during the process of acquiring the battery array image; The pose information includes: position information and orientation information; Based on the spatial information of the target detection box annotated with the abnormal battery, the image coordinates of the abnormal battery in the battery array image are determined; the spatial information includes: the center coordinates of the target detection box; Based on the pose information of the autonomous mobile device during the acquisition of the battery array image and the image coordinates of the abnormal battery in the battery array image, the position information of the abnormal battery in the set coordinate system is determined; Based on the location information of the abnormal battery in the set coordinate system, determine the location information of the abnormal battery in the data center; If the center coordinates are within the set target image coordinate range, and the confidence level that the abnormal battery is the target abnormal type is greater than or equal to the set confidence threshold, then the abnormal battery identification result is retained; or, if the center coordinates are not within the set target image coordinate range, and / or, the confidence level that the abnormal battery is the target abnormal type is less than the set confidence threshold, then the abnormal battery identification result is filtered. The target image coordinate range is used to filter battery arrays located in the central region of the battery array image.

5. The method of claim 4, wherein, The battery identification process for the battery array image includes: The battery array image is input into the battery detection model to obtain the spatial information of the target detection box for battery annotation in the battery array image; Based on the spatial information of the target detection box, extract the local image corresponding to the target detection box from the battery array image; Battery features are extracted from the local image to determine the battery features of the battery array image; The step of extracting battery features from the local image to determine the battery features of the battery array image includes: The local image is input into the feature extraction layer of the battery anomaly detection model to obtain the battery features of the local image.

6. The method of claim 5, wherein, The step of identifying anomalies in the target batteries contained in the battery array image based on the battery characteristics, to determine the abnormal batteries in the target batteries and the target anomaly type of the abnormal batteries, includes: The battery features are input into the anomaly recognition layer of the battery anomaly recognition model. In the anomaly recognition layer, the probability that the target battery contained in the local image belongs to multiple battery anomaly types is calculated based on the battery features. Based on the probability that the target battery belongs to multiple battery anomaly types, the abnormal batteries in the target battery and the target anomaly type of the abnormal batteries are determined.

7. The method of claim 4, wherein, The step of obtaining the pose information of the autonomous mobile device during the process of acquiring the battery array image includes: Acquire the acquisition points, as well as the height and orientation of the image acquisition device, recorded by the autonomous mobile device when acquiring images of the battery array; Based on the acquisition points recorded by the autonomous mobile device when acquiring the battery array image, as well as the height and orientation of the image acquisition device, the pose information of the autonomous mobile device during the acquisition of the battery array image is determined.

8. The method according to any one of claims 4-7, characterized in that, Also includes: The confidence level of the abnormal battery belonging to the target abnormal type is determined based on the probability that the abnormal battery belongs to the target abnormal type. Based on the target anomaly type of the abnormal battery, the confidence level that the abnormal battery belongs to the target anomaly type, and the location information of the abnormal battery in the data center, the anomaly identification result of the abnormal battery is determined; Output the anomaly identification result of the abnormal battery.

9. The method according to any one of claims 4-7, characterized in that, Also includes: The confidence level of the abnormal battery as the target abnormal type is determined based on the probability that the abnormal battery belongs to the target abnormal type.

10. The method of claim 8, wherein, Also includes: Based on the anomaly identification results of the target batteries contained in each frame of battery array image collected by the autonomous mobile device during the inspection of the data center, a battery inspection report of the data center is generated. Output the battery inspection report.

11. An autonomous mobile device, comprising: include: The mechanical body is equipped with an image acquisition device, a memory, and a processor. The memory is used to store computer programs; The image acquisition device is used to acquire images of the battery array during the movement of the autonomous mobile device along the battery array in the data center. The processor is coupled to the memory for executing the computer program to perform the steps of the method according to any one of claims 4-10.

12. A computing device, comprising: include: Memory, processor, and communication components; The memory is used to store computer programs; The processor is coupled to the memory and the communication component for executing the computer program to perform the steps of the method according to any one of claims 4-10.

13. A computer readable storage medium having stored thereon computer instructions, wherein: When the computer instructions are executed by one or more processors, the one or more processors are caused to perform the steps of the method according to any one of claims 4-10.

Images