Spatiotemporal multi-target behavior monitoring method, system and computing device

Abstract

The present invention discloses a spatiotemporal multi-target behavior monitoring method, system and computing device, belonging to the field of video image processing and computer vision technology. The method of the present invention includes: acquiring video data including multiple pedestrian targets; analyzing the video data to obtain key frames; using dual-stream task branches to extract features from the key frames to obtain human skeleton posture information and group motion characteristics; obtaining a spatial interaction information attention map based on human skeleton posture information, group motion characteristics and a graph convolutional network, and obtaining a group behavior recognition result based on the spatial interaction information attention map. The method of the present invention can deeply explore the spatial relationship and interaction pattern between targets, more accurately identify group behavior, improve the accuracy and reliability of multi-target behavior monitoring, effectively identify park life behavior, and promptly warn of abnormal behavior.

Description

Technical Field

[0001] The present invention relates to the field of video image processing and computer vision technology, and in particular to a spatiotemporal multi-target behavior monitoring method, system, computing device and computer-readable storage medium. Background Art

[0002] In recent decades, with the rise of neural networks, many methods have been developed to deal with behavior recognition problems. In visual human behavior analysis, it is necessary to analyze not only the spatial information of the detection target, but also its information in the time dimension, which is quite difficult. In addition, group behavior itself is complex and diverse, and video data is easily affected by perspective, occlusion, and interference from irrelevant personnel during the collection process, which is complex and diverse. Therefore, there is an urgent need for an efficient multi-target behavior monitoring method to achieve intelligent behavior control. Summary of the invention

[0003] The purpose of this application is to overcome the defects of the prior art and provide a spatiotemporal multi-target behavior monitoring method, system, computing device and computer-readable storage medium.

[0004] In a first aspect, the present application provides a spatiotemporal multi-target behavior monitoring method, comprising the following steps: Acquire video data including a plurality of pedestrian targets; Analyzing the video data to obtain key frame images; Using a dual-stream task branch to extract features from the key frame image to obtain human skeleton posture information and group motion characteristics; A spatial interaction information attention map is obtained based on the human skeleton posture information, the group action characteristics and the graph convolutional network, and a group behavior recognition result is obtained based on the spatial interaction information attention map.

[0005] Optionally, analyzing the video data frame by frame to obtain a key frame image includes: Define sliding windows; Sliding the sliding window along the time dimension on the video data to generate candidate interesting time intervals where actions may occur in the video data; The sliding window is slid within the time sequence interval to obtain the key frame image.

[0006] Optionally, the using a dual-stream task branch to perform feature extraction on the key frame image to obtain human skeleton posture information and group motion characteristics includes: Through the spatiotemporal multi-scale attention linkage mechanism, two task branches, static branch and dynamic branch, are established; Performing feature extraction on the key frame image through the static branch to obtain skeletal joints of the human body, and extracting the skeletal posture information of the human body based on the relationship between the skeletal joints and the human body; The optical flow and RGB information are refined and aggregated through the dynamic branch to obtain the group action characteristics.

[0007] Optionally, obtaining a spatial interaction information attention map based on the human skeleton posture information, the group motion characteristics and a graph convolutional network includes: Extracting and fusing the optical flow, the RGB information and the human skeleton posture information to obtain individual features; The individual features are input into a graph convolutional network for spatial interaction information modeling to obtain the spatial interaction information attention map.

[0008] Optionally, after obtaining a spatial interaction information attention map based on the human skeleton posture information, the group action characteristics and the graph convolutional network, and obtaining a group behavior recognition result based on the spatial interaction information attention map, the method further includes: The spatial interaction information attention map is input into a behavior recognition classifier to obtain and display the group behavior recognition result.

[0009] In a second aspect, the present application also provides a spatiotemporal multi-target behavior monitoring system, comprising: A data acquisition module, used for acquiring video data including multiple pedestrian targets; A preprocessing module, used for analyzing the video data to obtain key frame images; A processing module, used for extracting features from the key frame images using a dual-stream task branch to obtain human skeleton posture information and group motion characteristics; A recognition module is used to obtain a spatial interaction information attention map based on the human skeleton posture information, the group action characteristics and the graph convolution network, and obtain a group behavior recognition result based on the spatial interaction information attention map.

[0010] Optionally, the processing module and the recognition module are both based on a YOLOv8 network structure.

[0011] Optionally, the spatiotemporal multi-target behavior monitoring system further includes: A behavior recognition classifier is used to obtain and display the group behavior recognition result based on the spatial interaction information attention map.

[0012] In a third aspect, the present application also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the spatiotemporal multi-target behavior monitoring method as described in the first aspect are implemented.

[0013] In a fourth aspect, the present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the spatiotemporal multi-target behavior monitoring method as described in the first aspect.

[0014] The present application provides a spatiotemporal multi-target behavior monitoring method, system and computing device, which obtains video data including multiple pedestrian targets; analyzes the video data to obtain key frame images; uses a dual-stream task branch to extract features from the key frame images to obtain human skeleton posture information and group motion characteristics; obtains a spatial interaction information attention map based on the human skeleton posture information, the group motion characteristics and a graph convolutional network, and obtains a group behavior recognition result based on the spatial interaction information attention map; it can effectively identify campus life behaviors and provide timely warnings for abnormal behaviors, and has practical engineering application value.

[0015] In order to make the above features and advantages of the invention more obvious and easy to understand, embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technologies, the drawings required for use in the embodiments or the related technical descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 This is a flow chart of a spatiotemporal multi-target behavior monitoring method provided in one embodiment of the present application.

[0018] Figure 2 This is a flowchart of step S20 in the spatiotemporal multi-target behavior monitoring method provided in one embodiment of the present application.

[0019] Figure 3 This is a flowchart of step S30 in the spatiotemporal multi-target behavior monitoring method provided in one embodiment of the present application.

[0020] Figure 4 A schematic diagram of the structure of a spatiotemporal multi-target behavior monitoring system provided in one embodiment of the present application.

[0021] Figure 5 FIG. 4 is a diagram showing the internal structure of a computer device in yet another embodiment. DETAILED DESCRIPTION

[0022] In order to make the purpose and technical solution of the embodiment of the present application clearer, the technical solution of the embodiment of the present application will be clearly and completely described in conjunction with the drawings of the embodiment of the present application. Obviously, the described embodiment is a part of the embodiment of the present application, not all of the embodiments. Based on the described embodiment of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0023] In one embodiment, see Figure 1 The present application provides a spatiotemporal multi-target behavior monitoring method, which may include the following steps: S10~S50.

[0024] S10: Acquire video data including multiple pedestrian targets.

[0025] S20: Analyze the video data to obtain key frame images.

[0026] S30: Use the dual-stream task branch to extract features from the key frame image to obtain human skeleton posture information and group motion characteristics.

[0027] S40: A spatial interaction information attention map is obtained based on human skeleton posture information, group motion characteristics and a graph convolutional network, and a group behavior recognition result is obtained based on the spatial interaction information attention map.

[0028] S50: Inputting the spatial interaction information attention map into a behavior recognition classifier to obtain and display a group behavior recognition result.

[0029] In the spatiotemporal multi-target behavior monitoring method of the present application, by analyzing the video data to obtain key frame images, redundant information interference can be avoided and processing efficiency can be improved; using the dual-stream task branch to extract features from the key frame images, the behavioral characteristics of pedestrians in the video scene can be fully explored from different dimensions; using the graph convolutional network to model spatial interaction information to obtain a spatial interaction information attention map, it can deeply capture the relationship between individuals and enhance the reliability of group behavior recognition; the spatial interaction information attention map is input into the behavior recognition classifier to obtain and display the group behavior recognition results, which can more clearly display the monitoring results and realize multi-target behavior monitoring, which has important application value in security monitoring and other fields.

[0030] In step S10, refer to Figure 1 In step S10, video data including multiple pedestrian targets is obtained.

[0031] As an example, video data including multiple pedestrian targets or continuously shot image data may be obtained from a video shooting device such as a camera or a monitoring device.

[0032] In step S20, refer to Figure 1 In step S20, the video data is analyzed to obtain a key frame image.

[0033] As an example, see Figure 2 , step S20 may include the following steps: S201~S205.

[0034] S201: Define a sliding window.

[0035] S202: Slide the sliding window along the time dimension on the video data to generate candidate interesting time intervals where actions may occur in the video data.

[0036] S203: Slide the sliding window within the time interval to obtain a key frame image.

[0037] As an example, read the video data and use the temporal attention mechanism to extract features from the video to obtain timing information. Define a sliding window, the length of which can be set according to actual needs. The sliding window slides along the time dimension on the video and finds the candidate interesting time intervals where actions may occur in the video based on the timing information, that is, the time intervals where pedestrians appear in the video. Further, input the time intervals into the sliding window to obtain key frame images.

[0038] As an example, in order to accurately divide the boundary area of ​​the time series interval, a rolling window generation function rolling_window is defined, and the sliding window size is dynamically adjusted by setting the offset. Specifically, for the boundary area, the offset is iterated multiple times to dynamically adjust the size of the sliding window to fit different time series intervals, divide the boundary more accurately, and remove videos without pedestrians. The rolling window generation function can be expressed by the following formula: in, r is the learning rate, is the gradient, is the updated size of the window, β is the offset.

[0039] As an example, by dynamically adjusting the sliding window size, the boundaries of the time series interval can be determined more accurately. In video analysis, the start and end points of many behaviors are not uniform, and there are often fuzzy transition areas. This dynamic adjustment mechanism can flexibly adapt the boundaries according to the actual changes in pedestrian behavior characteristics, avoiding the problem of inaccurate boundary division caused by fixed window size. It can also significantly reduce the amount of data processing, so that computing resources can be more concentrated on the part with pedestrian behavior information, improve the efficiency of the entire analysis process, and avoid the interference of pedestrian-free areas on behavior analysis. Through the dynamic adjustment mechanism based on the offset, whether in high-dynamic scenes where pedestrians move quickly and interact frequently, or in low-dynamic scenes where pedestrians move slowly and have relatively simple behaviors, the window size can be automatically adjusted to adapt to the temporal changes of behavior, ensuring that accurate time series intervals can be effectively extracted in various complex and changeable practical application scenarios.

[0040] As an example, the experimental environment for building the rolling window generation function can use the Windows operating system or other operating systems that can be programmed, Python can be used as the main programming language to implement algorithms and models, PyTorch can be used as a deep learning framework to build and train neural network models, and libraries such as Pandas can be used to complete data loading, processing and conversion. Such an experimental environment provides a convenient and efficient platform that is conducive to the construction of algorithms and systems.

[0041] In step S30, refer to Figure 1 In step S30, the dual-stream task branch is used to extract features from the key frame images to obtain human skeleton posture information and group motion characteristics.

[0042] As an example, see Figure 3 , step S30 may include the following steps: S301~S303.

[0043] S301: Through the spatiotemporal multi-scale attention linkage mechanism, two task branches, static branch and dynamic branch, are established.

[0044] S302: extracting features from the key frame image through a static branch to obtain the skeletal joints of the human body, and extracting the skeletal posture information of the human body based on the relationship between the skeletal joints and the human body.

[0045] S303: Refine and aggregate the optical flow and RGB information through dynamic branches to obtain group action characteristics.

[0046] As an example, in step S301, two parallel tasks, a static branch and a dynamic branch, are established through the spatiotemporal multi-scale attention linkage mechanism. The spatiotemporal information is obtained from the two dimensions of time and space. The static branch focuses on the posture contour, and the dynamic branch tracks the motion changes. By aggregating static postures and dynamic characteristics, the spatiotemporal information of the key frame human body can be obtained.

[0047] As an example, the spatiotemporal information may include scene appearance of the input video data, postures of group members, optical flow of member motion, time series dependencies between frames, etc.

[0048] As an example, in step S302, the joint points of the human skeleton in the key frame are found through the posture attention mechanism, and the human skeleton posture information is obtained according to the relationship between the joint points and the human body. After multiple iterations, all the individual information in the key frame is found. By capturing the joint points and focusing on the key posture parts through static branches, individual information can be quickly located and extracted.

[0049] As an example, in step S303, the dynamic branch extracts features from the key frame and the previous and next frames to obtain individual temporal characteristics, and refines and aggregates the optical flow and RGB information of the image to obtain individual motion characteristics. After multiple iterations, the motion characteristics of all individuals in the key frame are obtained. Optical flow is the visual apparent motion formed by the displacement of the pixel intensity pattern in the image due to the movement of the target. It reflects the motion vector of the pixel between adjacent frames and can capture the direction and speed of the target movement. RGB information is the red, green, and blue primary color data of the image, which contains the appearance characteristics of the target and the context information of the scene. The dynamic branch network obtains pixel motion information through optical flow, obtains morphology and environmental information through RGB information, fuses the optical flow and RGB information according to certain weight parameters, aggregates important individual information, removes unimportant information, and thus reduces the network burden.

[0050] As an example, the key frame data is simultaneously input into the dynamic branch network and the static branch network for iterative updating to obtain the individual information and individual action information of all individuals in the key frame image. The two pieces of information are respectively assigned a weight for dynamic fusion to obtain the group action characteristics.

[0051] As an example, the static branch and the dynamic branch can be implemented based on the YOLOv8 network structure.

[0052] In step S40, refer to Figure 1 In step S40, a spatial interaction information attention map is obtained based on human skeleton posture information, group motion characteristics and a graph convolutional network, and a group behavior recognition result is obtained based on the spatial interaction information attention map.

[0053] As an example, based on the motion information obtained by refining optical flow and RGB information, an exclusive tracking trajectory is established for each individual. The target tracking algorithm determines the individual's new position and state based on the correlation of trajectory feature points between consecutive frames and the matching of appearance features, and constructs a coherent spatiotemporal motion trajectory to ensure that the individual trajectory is accurately continued without loss or mismatch.

[0054] As an example, human skeleton posture information and group motion characteristics are constructed as graph structure data and input into the graph convolutional network. The graph convolutional network operates on graph nodes (individuals) and edges (potential interactions between individuals) through convolution kernels, aggregates the features of neighboring nodes, and then updates the node representation to mine the spatial interaction information between individuals, transforming individual features from isolated individuals to structured spatial interaction information in group association contexts.

[0055] As an example, adaptively adjustable weight parameters are assigned to each channel of the graph convolutional network, and the weights of the spatial interaction information channels are dynamically adjusted to adapt to the dynamics of the network environment and enhance the network's spatial relationship reasoning capabilities.

[0056] As an example, the loss function can be formulated as follows: in, is the true value, is the predicted value, M is the number of samples, is the weight, is the label, MSE is the mean square error, which is the mean of the square of the error between the predicted value and the true value, and MAE is the mean absolute error, which is the average of the absolute error between the predicted value and the true value. If the model is too sensitive to outliers, reduce , increase the MAE ratio, if you want to prevent the generalization of the model, you can soften the label, you can Change to the following formula: in, is a constant, which can be 0.1. For the changed label, This is the label before the change.

[0057] As an example, the contribution of group feature extraction is quantitatively ranked according to the updated weights, and the key weights are retained to maintain or increase their proportions. The weights with low contribution are linearly or exponentially decayed, such as setting the attenuation factor to reduce the weight value according to the iteration rounds, thereby reducing the impact of weakly correlated features, thereby simplifying the model structure, focusing on the core behavior representation and preventing overfitting.

[0058] In step S50, refer to Figure 1 In step S50, the spatial interaction information attention map is input into a behavior recognition classifier to obtain and display a group behavior recognition result.

[0059] As an example, the obtained spatial interaction information attention map is input into the pre-trained behavior recognition classifier. According to the interaction relationship between group members, group behavior can be divided into two categories: group behavior without interaction relationship and group behavior based on interaction relationship. Finally, the classification result is output, and the recognition result is output in the form of a visual image, annotated with group labels and confidence, to generate the final accurate and intuitive group behavior for user decision-making application.

[0060] As an example, the behavior recognition classifier may be a support vector machine, a deep neural network classifier, SORT, etc.

[0061] In the spatiotemporal multi-target behavior monitoring method of this application, key frames are found through the time attention mechanism, joint points are captured based on the posture attention mechanism, dual-stream task branches are established through the spatiotemporal multi-scale attention linkage mechanism, human skeleton posture information is extracted through static branches, group motion characteristics are refined and aggregated through dynamic branches, and finally individual and group behavior recognition classifiers are used to complete the monitoring of multi-target behaviors. This application method can effectively identify life behaviors in the park and provide timely warnings for abnormal behaviors, and has practical engineering application value.

[0062] In another embodiment, see Figure 4 The present application also provides a spatiotemporal multi-target behavior monitoring system which may include: a data acquisition module 10, a preprocessing module 20, a processing module 30, and a recognition module 40. The data acquisition module 10 is used to acquire video data including multiple pedestrian targets; the preprocessing module 20 is used to analyze the video data to obtain key frame images; the processing module 30 is used to extract features from the key frame images using a dual-stream task branch to obtain human skeleton posture information and group motion characteristics; the recognition module 40 is used to obtain a spatial interaction information attention map based on human skeleton posture information, group motion characteristics and a graph convolutional network, and obtain a group behavior recognition result based on the spatial interaction information attention map.

[0063] As an example, the processing module 30 and the recognition module 40 are both based on the YOLOv8 network structure.

[0064] As an example, the spatiotemporal multi-target behavior monitoring system may further include a behavior recognition classifier 50, which is used to obtain and display group behavior recognition results based on the spatial interaction information attention map.

[0065] In the above-mentioned spatiotemporal multi-target behavior monitoring system, the data acquisition module 10 acquires video data including multiple pedestrian targets, the pre-processing module 20 finds the key frame image, the processing module 30 obtains the human skeleton posture information and group motion characteristics, the recognition module 40 obtains the group behavior recognition result, and the group behavior recognition result is displayed through the behavior recognition classifier 50. It helps to fully and deeply grasp the activity status of multiple targets in the monitoring scene, effectively identify the life behavior in the park, and timely warn of abnormal behavior. It provides strong support for security, park management, etc., such as timely preventing the invasion of dangerous areas or handling unexpected situations of personnel, effectively protecting the safety of life and property of personnel, and enhancing environmental safety and stability.

[0066] In another embodiment, the present invention also provides a computer device, which may include a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above-mentioned spatiotemporal multi-target behavior monitoring method when executing the computer program.

[0067] As an example, the computer device may be a server, and its internal structure diagram may be as follows: Figure 5 As shown. The computer device includes a processor, a memory, a network interface and a database connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data such as regional parameters and deployment strategies. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the spatiotemporal multi-target behavior monitoring method in the above embodiment is implemented.

[0068] Those skilled in the art will understand that Figure 5 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0069] In yet another embodiment, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements all the steps of the spatiotemporal multi-target behavior monitoring method in the above-mentioned embodiment.

[0070] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0071] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0072] Although the present application has been disclosed as above with the embodiments, it is not intended to limit the present application. Any person with ordinary knowledge in the technical field can make some changes and modifications without departing from the spirit and scope of the present application. Therefore, the scope of protection of the present application shall be determined by the scope of the attached patent application.

Claims

1. A spatiotemporal multi-target behavior monitoring method, characterized in that: The following steps are involved: Acquire video data including a plurality of pedestrian targets; Analyzing the video data to obtain key frame images; Using a dual-stream task branch to extract features from the key frame image to obtain human skeleton posture information and group motion characteristics; A spatial interaction information attention map is obtained based on the human skeleton posture information, the group action characteristics and the graph convolutional network, and a group behavior recognition result is obtained based on the spatial interaction information attention map.

2. The spatiotemporal multi-target behavior monitoring method according to claim 1, characterized in that: The step of analyzing the video data frame by frame to obtain a key frame image includes: Define sliding windows; Sliding the sliding window along the time dimension on the video data to generate candidate interesting time intervals where actions may occur in the video data; The sliding window is slid within the time sequence interval to obtain the key frame image.

3. The spatiotemporal multi-target behavior monitoring method according to claim 1, characterized in that: The using of the dual-stream task branch to extract features from the key frame image to obtain human skeleton posture information and group motion characteristics includes: Through the spatiotemporal multi-scale attention linkage mechanism, two task branches, static branch and dynamic branch, are established; Performing feature extraction on the key frame image through the static branch to obtain skeletal joints of the human body, and extracting the skeletal posture information of the human body based on the relationship between the skeletal joints and the human body; The optical flow and RGB information are refined and aggregated through the dynamic branch to obtain the group action characteristics.

4. The spatiotemporal multi-target behavior monitoring method according to claim 1, characterized in that: The spatial interaction information attention map is obtained based on the human skeleton posture information, the group action characteristics and the graph convolution network, including: Extracting and fusing the optical flow, the RGB information and the human skeleton posture information to obtain individual features; The individual features are input into a graph convolutional network for spatial interaction information modeling to obtain the spatial interaction information attention map.

5. The spatiotemporal multi-target behavior monitoring method according to any one of claims 1 to 4, characterized in that: After obtaining a spatial interaction information attention map based on the human skeleton posture information, the group action characteristics and the graph convolution network, and obtaining a group behavior recognition result based on the spatial interaction information attention map, the method further includes: The spatial interaction information attention map is input into a behavior recognition classifier to obtain and display the group behavior recognition result.

6. A spatiotemporal multi-target behavior monitoring system, characterized in that: The spatiotemporal multi-target behavior monitoring system comprises: A data acquisition module, used for acquiring video data including multiple pedestrian targets; A preprocessing module, used for analyzing the video data to obtain key frame images; A processing module, used for extracting features from the key frame images using a dual-stream task branch to obtain human skeleton posture information and group motion characteristics; A recognition module is used to obtain a spatial interaction information attention map based on the human skeleton posture information, the group action characteristics and the graph convolution network, and obtain a group behavior recognition result based on the spatial interaction information attention map.

7. The spatiotemporal multi-target behavior monitoring system according to claim 6, characterized in that: The processing module and the recognition module are both based on the YOLOv8 network structure.

8. The spatiotemporal multi-target behavior monitoring system according to claim 6 or 7, characterized in that: The spatiotemporal multi-target behavior monitoring system also includes: A behavior recognition classifier is used to obtain and display the group behavior recognition result based on the spatial interaction information attention map.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images