Fine-grained atmospheric pollution distribution speculation method and device and storage medium
By combining the self-attention convolutional network and the bidirectional optical flow feedforward layer, the spatiotemporal change mapping of air pollutants from coarse-grained to fine-grained is achieved, which solves the problem of uneven distribution of air quality monitoring stations and improves the accuracy and stability of air pollutant distribution prediction.
Patent Information
- Application Number
- CN202510660903.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-22
- Publication Date
- 2025-09-19
AI Technical Summary
In existing technologies, the distribution of air quality monitoring stations within cities is uneven, making it difficult to obtain regional characteristic data on air pollution between multiple cities and provinces. This results in a lack of fine-grained distribution information at the macro level in air pollution monitoring data.
A self-attention convolutional network and a bidirectional optical flow feedforward layer are used to construct a spatiotemporal super-resolution proxy task. By mapping the spatiotemporal changes from coarse-grained pollution distribution maps to fine-grained pollution distribution maps, and utilizing the correlation between historical data and future predictions, accurate inference of air pollutants can be achieved.
It improves the accuracy and stability of air pollutant distribution prediction, can more accurately capture the changing patterns of air pollutants at different locations and times, and solves the problem of accurate inference of the fine-grained distribution of air pollutants.
Smart Images
Figure CN120671810A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of environmental monitoring, and in particular to a fine-grained atmospheric pollution distribution inference method, device and storage medium. Background Art
[0002] Monitoring the distribution of fine particulate matter air pollution is crucial for urban planning. Air pollution often exhibits regional characteristics, stemming from both the cross-regional transport of pollutants caused by atmospheric circulation and the irrational industrial layout and extensive economic development models in urban areas. To effectively manage and control air pollution sources, it is essential to pay attention to the distribution of fine particulate matter pollution monitoring data.
[0003] However, in current air pollution monitoring, most monitoring stations are concentrated in urban centers and around industrial areas, while relatively few are located in suburban and rural areas. This highly uneven distribution of air quality monitoring stations within cities makes it difficult to obtain data on the regional characteristics of air pollution at a more macro level, such as across multiple cities and provinces. Summary of the Invention
[0004] In order to solve the above technical problems, the present invention provides a fine-grained atmospheric pollution distribution estimation method.
[0005] In order to solve the above technical problems, the present invention adopts the following technical solutions:
[0006] In a first aspect, the present invention provides a fine-grained atmospheric pollution distribution estimation method, which uses a trained estimation model to estimate a fine-grained pollution distribution map based on a coarse-grained pollution distribution map. The estimation model training process includes:
[0007] Based on the historical pollutant distribution data of a given area, the pollution distribution map is divided into a fine-grained pollution distribution map grid sequence, and a coarse-grained pollution distribution map grid sequence is calculated according to the set upsampling factor;
[0008] The grid sequence of coarse-grained pollution distribution maps is input into the self-attention convolutional network to extract the spatial correlation characteristics of the spatiotemporal diffusion of pollution distribution; the spatial correlation characteristics of the spatiotemporal diffusion of pollution distribution are input into the bidirectional optical flow feedforward layer, and the temporal difference characteristics of the pollutant diffusion movement are learned to obtain the time-related characteristics of the pollutant diffusion; a spatiotemporal super-resolution proxy task inference module is constructed to realize the spatiotemporal change mapping of the coarse-grained distribution of pollutants to the fine-grained distribution based on the time-related characteristics of the pollutant diffusion, thereby realizing the training of the inference model.
[0009] In one embodiment, the method of dividing the pollution distribution map into a fine-grained pollution distribution map grid sequence based on historical pollutant distribution data of a given area, and calculating the coarse-grained pollution distribution map grid sequence based on a set upsampling factor, specifically includes:
[0010] In the designated target pollution area, the pollution distribution map is evenly divided into an H×W fine-grained pollution distribution map grid sequence Y(1), Y(2), Y(3), …, Y(t) according to the specified spatial grid size; Y(t) represents the pollutant value of the fine-grained pollution distribution map grid at a specific location at the t-th moment; the average pollutant value of the fine-grained pollution distribution map grid is calculated according to the set upsampling factor r, and the construction size is The coarse-grained pollution distribution map grid sequence X=X(1),X(2),X(3),…,X(t), where X(t) represents the pollutant value of the coarse-grained pollution distribution map grid at a specific position at the t-th moment, where the pollutant value of the coarse-grained pollution distribution map grid is equal to the average value of the pollutants in the corresponding fine-grained pollution distribution map grid.
[0011] In one embodiment, the step of inputting the coarse-grained pollution distribution grid sequence into a self-attention convolutional network to extract the spatial correlation features of the spatiotemporal diffusion of pollution distribution specifically includes:
[0012] The coarse-grained pollution distribution map grid sequence X is input into three independent convolutional networks to extract spatial information. The expansion operation is used to extract the three-dimensional cube feature blocks of the local feature map with a sliding window size of 1; the obtained three-dimensional cube feature blocks are reconstructed into a one-dimensional feature vector to obtain the query vector Q, key vector K and value vector V:
[0013] Q=f unfold (W Q ×X);
[0014] K=f unfold ((W K ×X);
[0015] V=f unfokd ((W V ×X);
[0016] Among them, Q, K, and V represent the query vector, key vector, and value vector in the self-attention mechanism respectively, and W Q 、W K 、W V is the corresponding weight matrix, f unfold Indicates the expansion operation;
[0017] The spatial correlation characteristics of the pollution distribution and spatiotemporal diffusion are obtained using the sliding window of the three-dimensional cube feature block. a tt en ti on :
[0018]
[0019] Among them, ffold is the folding operation, h is the number of heads in the self-attention mechanism, is the attention output weight matrix, which is used to map the attention output of each head of the self-attention mechanism to the final output space; φ and σ1 are activation functions used to introduce nonlinear transformations; are the weight matrices corresponding to the i-th attention head.
[0020] In one embodiment, the spatial correlation features of the spatiotemporal diffusion of the pollution distribution are input into the bidirectional optical flow feedforward layer, and temporal difference feature learning is performed on the pollutant diffusion movement to obtain the temporal correlation features of the pollutant diffusion, specifically including:
[0021] The spatial correlation characteristics of the pollution distribution and spatiotemporal diffusion are aggregated using the pollutant optical flow differential information and time-varying data to obtain the forward propagation characteristics of the pollutants. and backpropagation features and
[0022]
[0023] in, and They represent the forward optical flow change information and the reverse optical flow change information of the pollutant distribution at the tth moment respectively; ω is the function used to merge the optical flow information with the original image sequence;
[0024] The time-related features of pollutant diffusion f are obtained by fusion of bidirectional propagation features of optical flow output (X):
[0025]
[0026] Among them, ρ is the fusion operation, ResidualConv is the convolution residual block, φ represents the activation function, and f attention Represents the spatial correlation characteristics of the temporal and spatial diffusion of pollution distribution.
[0027] In one embodiment, the construction of the spatiotemporal super-resolution proxy task inference module realizes spatiotemporal change mapping from coarse-grained distribution to fine-grained distribution of pollutants based on the time-related characteristics of pollutant diffusion, specifically including:
[0028] According to the specified upsampling factor r, the time-dependent characteristics f of the pollutant diffusion are output Perform upsampling(·) to obtain the inferred fine-grained pollution distribution map
[0029]
[0030] Through the mean square error loss function Optimizing the inference model:
[0031]
[0032] T represents the total number of moments in the grid sequence of the fine-grained pollution distribution map, represents the pollutant value of the fine-grained pollution distribution grid at a specific location at the t-th moment; Y(t) represents the pollutant value of the actual fine-grained pollution distribution grid at a specific location at the t-th moment.
[0033] In a second aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the processor implements the steps of the method of any one embodiment of the first aspect when executing the computer program.
[0034] In a third aspect, the present invention provides 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 of any one embodiment of the first aspect are implemented.
[0035] Compared with the prior art, the beneficial technical effects of the present invention are:
[0036] This paper proposes a method for estimating fine-grained atmospheric pollution distribution. By introducing a three-dimensional self-attention convolutional layer, a bidirectional optical flow feedback layer, and a coarser-to-coarser-grained proxy task, the model simultaneously considers local and global information in the spatiotemporal domain, thereby more accurately capturing the changing patterns of air pollutants at different locations and over time. Furthermore, the introduction of the bidirectional optical flow feedback layer enables the model to effectively leverage the correlation between historical data and future forecasts, improving the accuracy and stability of the model's predictions and thus solving the problem of accurately inferring the fine-grained distribution of atmospheric pollutants. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] Figure 1 Flowchart of a method in an embodiment of the present invention.
[0038] Figure 2 Schematic diagram of the difference between the true value and the estimated value of the pollution distribution in an embodiment of the present invention. DETAILED DESCRIPTION
[0039] A preferred embodiment of the present invention will be described in detail below with reference to the accompanying drawings.
[0040] like Figure 1 As shown, the present invention provides a fine-grained atmospheric pollution distribution inference method, which infers a fine-grained pollution distribution map based on a coarse-grained pollution distribution map through a trained inference model; the inference model training process includes the following steps:
[0041] S1, based on the historical pollutant distribution data of a given area, divides the pollution distribution map into a fine-grained pollution distribution map grid sequence, and calculates the coarse-grained pollution distribution map grid sequence according to the set upsampling factor.
[0042] S2, the coarse-grained pollution distribution map grid sequence is input into the self-attention convolutional network to extract the spatial correlation features of the spatiotemporal diffusion of pollution distribution.
[0043] S3, inputting the spatial correlation features of the spatiotemporal diffusion of the pollution distribution into the bidirectional optical flow feedforward layer, performing time difference feature learning on the pollutant diffusion movement, and obtaining the time correlation features of the pollutant diffusion.
[0044] S4, constructs a spatiotemporal super-resolution agent task inference module to achieve spatiotemporal change mapping from coarse-grained distribution to fine-grained distribution of pollutants based on the time-related characteristics of pollutant diffusion.
[0045] By introducing a three-dimensional self-attention convolutional layer, a bidirectional optical flow feedback layer, and a coarser-to-coarser-grained proxy task, this model can simultaneously consider local and global information in the spatiotemporal domain, thereby more accurately capturing the changing patterns of air pollutants at different locations and over time. Furthermore, the introduction of the bidirectional optical flow feedback layer enables the model to effectively leverage the correlation between historical data and future predictions, improving the accuracy and stability of the model's predictions.
[0046] In one embodiment, step S1 divides the pollution distribution map into a fine-grained pollution distribution map grid sequence based on historical pollutant distribution data of a given area, and calculates a coarse-grained pollution distribution map grid sequence based on a set upsampling factor, specifically including:
[0047] In the designated target pollution area, the pollution distribution map is evenly divided into an H×W fine-grained pollution distribution map grid sequence Y(1), Y(2), Y(3), …, Y(t) according to the specified spatial grid size; Y(t) represents the pollutant value of the fine-grained pollution distribution map grid at a specific location at the t-th moment; the average pollutant value of the fine-grained pollution distribution map grid is calculated according to the set upsampling factor r, and the construction size is The coarse-grained pollution distribution map grid sequence X=X(1),X(2),X(3),…,X(t), where X(t) represents the pollutant value of the coarse-grained pollution distribution map grid at a specific position at the t-th moment, where the pollutant value of the coarse-grained pollution distribution map grid is equal to the average value of the pollutants in the corresponding fine-grained pollution distribution map grid.
[0048] A "pollution map" is a graphic or image that depicts the distribution of atmospheric pollutants. This image uses data visualization to illustrate the spatial distribution and concentration of pollutants within a specific area. Typically, such maps are based on real-time data from environmental monitoring stations or other relevant data sources.
[0049] The coarse-grained pollution distribution map provides pollution data for a larger area or a wider range, while the fine-grained pollution distribution map provides pollution data for a smaller area or a finer range. The coarse-grained pollution distribution map and the fine-grained pollution distribution map are relative concepts. It can be understood that the pollution data for an area larger than the set range is called a coarse-grained pollution distribution map, and the pollution data for an area smaller than the set range is called a fine-grained pollution distribution map.
[0050] In one embodiment, step S2 of inputting the coarse-grained pollution distribution grid sequence into the self-attention convolutional network to extract the spatial correlation features of the spatiotemporal diffusion of pollution distribution specifically includes:
[0051] The coarse-grained pollution distribution map grid sequence X is input into three independent convolutional networks to extract spatial information, and the sliding local feature map is extracted using the expansion operation. The stride of the expansion operation is 1, and the shape of the local feature map is T×W p ×H p .
[0052] The obtained three-dimensional cube feature block (3D-patch) is reconstructed into a one-dimensional feature vector to obtain the query vector Q and key vector K. The similarity matrix is calculated using the dot product and aggregated with the value vector V into a feature graph.
[0053] Q=f unfold (W Q ×X);
[0054] K=f unfold ((W K ×X);
[0055] V=f unfold ((W V ×X);
[0056] Among them, Q, K, and V represent the query vector, key vector, and value vector in the self-attention mechanism, respectively, which are obtained by linear transformation of the coarse-grained pollution distribution grid sequence X, where W Q 、W K 、W V is the corresponding weight matrix, f unfold Represents an expand operation.
[0057] Use 3D-patch sliding window to obtain the spatial correlation characteristics of pollution distribution and spatiotemporal diffusion f attention:
[0058]
[0059] Among them, f fold is the folding operation, h is the number of heads in the self-attention mechanism, is the attention output weight matrix, which is used to map the attention output of each head of the self-attention mechanism to the final output space; φ and σ1 are activation functions used to introduce nonlinear transformations.
[0060] In one embodiment, step S3 inputs the spatial correlation features of the spatiotemporal diffusion of the pollution distribution into the bidirectional optical flow feedforward layer, performs temporal difference feature learning on the pollutant diffusion movement, and obtains the temporal correlation features of the pollutant diffusion, specifically including:
[0061] The spatial correlation characteristics of the pollution distribution and spatiotemporal diffusion are aggregated using the pollutant optical flow differential information and time-varying data to obtain the forward propagation characteristics of the pollutants. and backpropagation features and
[0062]
[0063] Here, ω is the function used to merge the optical flow information with the original image sequence.
[0064] and They represent the forward optical flow change information and reverse optical flow change information of the pollutant distribution at the tth moment respectively:
[0065]
[0066] V t It represents the direction of optical flow movement at the tth moment, and s represents the calculation of optical flow similarity function.
[0067] The time-related features of pollutant diffusion f are obtained by fusion of bidirectional propagation features of optical flow output (X):
[0068]
[0069] Among them, ρ is the fusion operation, ResidualConv is the convolution residual block, φ represents the activation function, and f attention Represents the spatial correlation characteristics of the temporal and spatial diffusion of pollution distribution.
[0070] In one embodiment, the construction of the spatiotemporal super-resolution proxy task inference module in step S4 implements spatiotemporal change mapping from coarse-grained distribution to fine-grained distribution of pollutants based on the time-related characteristics of pollutant diffusion, specifically including:
[0071] According to the specified upsampling factor r, the time-dependent characteristics f of the pollutant diffusion are output Upsampling is performed to obtain the inferred fine-grained pollution distribution map
[0072]
[0073] Through the mean square error loss function Optimizing the inference model:
[0074]
[0075] represents the pollutant value of the fine-grained pollution distribution grid at a specific location at the t-th moment; Y(t) represents the pollutant value of the actual fine-grained pollution distribution grid at a specific location at the t-th moment.
[0076] The training set was partitioned into 100%, 80%, and 40% partitions. Table 1 shows that the inference model of the present invention outperforms other models (Methods, Mean, HA, Urban FM, EDVR, basicVSR, IconVSR, basicVSR++, and MANA) in all of the above partition ratios. In the 100% partition, the inference model designed by the present invention outperforms the other models by 2.6%, 3.05%, and 6.36% in root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE). In the 40% partition, the inference model of the present invention also outperforms the second-best model by 3.86%, 3.76%, and 12.18% in RMSE, MAE, and MAPE.
[0077] Table 1, average RMSE, MAE, and MAPE of experiments on datasets divided by different data sizes:
[0078]
[0079]
[0080] Table 2 shows the comparison of the training time and parameters between the inference model of the present invention (AirSTFM) and other models (Urban FM, EDVR, basicVSR, IconVSR, basicVSR++, MANA). In the comparison with the IconVSR model, its inference performance ranks second, but the training time is twice as long as that of the model of this method.
[0081] Table 2, model training parameters and training time:
[0082]
[0083] Figure 2 The difference between the estimated values and the true values of the inference model (AirSTFM) of the present invention and other models (Mean, HA, Urban-FM, MANA, BasicVSR, BasicVSR++, IconVSR) is shown. The higher the brightness, the greater the error. It can be seen that most of them are distributed in the middle reaches of a certain river and coastal areas. This is because the cities in the middle and lower reaches of a certain river, as heavy industrial cities, produce more serious pollution, and the emissions of atmospheric pollutants change rapidly, which makes the inference more difficult and leads to larger errors.
[0084] It should be understood that although the steps in the flowcharts of the accompanying drawings are shown in sequence as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least some of the steps in the flowcharts of the accompanying drawings may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times. The order of execution of these steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a portion of the steps or stages in other steps.
[0085] In one embodiment, the present invention provides a computer device comprising a processor, a memory, and a network interface connected via a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises 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 computer program in the non-volatile storage medium. The database of the computer device is configured to store data used in the above-described method. The network interface of the computer device is configured to communicate with an external terminal via a network connection. The computer program, when executed by the processor, implements the above-described method.
[0086] In an exemplary embodiment, a computer-readable storage medium including instructions, such as a memory including instructions, is also provided. The instructions are executable by a processor to perform the above method. The storage medium can be a computer-readable storage medium, such as a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, or the like.
[0087] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. It is intended that all variations within the meaning and range of equivalents of the claims be embraced herein, and any reference signs in the claims should not be construed as limiting the claims to which they relate.
[0088] In addition, it should be understood that although this specification is described in terms of implementation methods, not every implementation method contains only one independent technical solution. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.
Claims
1. A fine-grained atmospheric pollution distribution estimation method, characterized in that: The trained inference model is used to infer the fine-grained pollution distribution map based on the coarse-grained pollution distribution map. The training process of the inference model includes: Based on the historical pollutant distribution data of a given area, the pollution distribution map is divided into a fine-grained pollution distribution map grid sequence, and a coarse-grained pollution distribution map grid sequence is calculated according to the set upsampling factor; The grid sequence of coarse-grained pollution distribution maps is input into the self-attention convolutional network to extract the spatial correlation characteristics of the spatiotemporal diffusion of pollution distribution; the spatial correlation characteristics of the spatiotemporal diffusion of pollution distribution are input into the bidirectional optical flow feedforward layer, and the temporal difference characteristics of the pollutant diffusion movement are learned to obtain the time-related characteristics of the pollutant diffusion; a spatiotemporal super-resolution proxy task inference module is constructed to realize the spatiotemporal change mapping of the coarse-grained distribution of pollutants to the fine-grained distribution based on the time-related characteristics of the pollutant diffusion, thereby realizing the training of the inference model.
2. The fine-grained atmospheric pollution distribution estimation method according to claim 1, characterized in that: The method of dividing the pollution distribution map into a fine-grained pollution distribution map grid sequence based on the historical pollutant distribution data of a given area and calculating the coarse-grained pollution distribution map grid sequence according to the set upsampling factor specifically includes: In the designated target pollution area, the pollution distribution map is evenly divided into an H×W fine-grained pollution distribution map grid sequence Y(1), Y(2), Y(3), …, Y(t) according to the specified spatial grid size; Y(t) represents the pollutant value of the fine-grained pollution distribution map grid at a specific location at the t-th moment; the average pollutant value of the fine-grained pollution distribution map grid is calculated according to the set upsampling factor r, and the construction size is The coarse-grained pollution distribution map grid sequence X=X(1),X(2),X(3),…,X(t), where X(t) represents the pollutant value of the coarse-grained pollution distribution map grid at a specific position at the t-th moment, where the pollutant value of the coarse-grained pollution distribution map grid is equal to the average value of the pollutants in the corresponding fine-grained pollution distribution map grid.
3. The fine-grained atmospheric pollution distribution estimation method according to claim 1, characterized in that: The coarse-grained pollution distribution map grid sequence is input into the self-attention convolutional network to extract the spatial correlation features of the spatiotemporal diffusion of pollution distribution, specifically including: The coarse-grained pollution distribution map grid sequence X is input into three independent convolutional networks to extract spatial information. The expansion operation is used to extract the three-dimensional cube feature blocks of the local feature map with a sliding window size of 1; the obtained three-dimensional cube feature blocks are reconstructed into a one-dimensional feature vector to obtain the query vector Q, key vector K and value vector V: Q=f unfold (W Q ×X); K=f unfold ((W K ×X); V=f unfold ((w V ×X); Among them, Q, K, and V represent the query vector, key vector, and value vector in the self-attention mechanism respectively, and W Q 、W K 、W V is the corresponding weight matrix, f unfold Indicates the expansion operation; The spatial correlation characteristics of the pollution distribution and spatiotemporal diffusion are obtained using the sliding window of the three-dimensional cube feature block. atten ti on : Among them, f fold is the folding operation, h is the number of heads in the self-attention mechanism, is the attention output weight matrix, which is used to map the attention output of each head of the self-attention mechanism to the final output space; φ and σ1 are activation functions used to introduce nonlinear transformations; are the weight matrices corresponding to the i-th attention head.
4. The fine-grained atmospheric pollution distribution estimation method according to claim 1, characterized in that: The spatial correlation features of the spatiotemporal diffusion of the pollution distribution are input into the bidirectional optical flow feedforward layer, and the time difference feature learning of the pollutant diffusion movement is performed to obtain the time correlation features of the pollutant diffusion, specifically including: The spatial correlation characteristics of the pollution distribution and spatiotemporal diffusion are aggregated using the pollutant optical flow differential information and time-varying data to obtain the forward propagation characteristics of the pollutants. and backpropagation features and in, and They represent the forward optical flow change information and the reverse optical flow change information of the pollutant distribution at the tth moment respectively; ω is the function used to merge the optical flow information with the original image sequence; The time-related features of pollutant diffusion f are obtained by fusion of bidirectional propagation features of optical flow output (X): Among them, ρ is the fusion operation, ResidualConv is the convolution residual block, φ represents the activation function, and f attention Represents the spatial correlation characteristics of the temporal and spatial diffusion of pollution distribution.
5. The fine-grained atmospheric pollution distribution estimation method according to claim 1, characterized in that: The construction of the spatiotemporal super-resolution agent task inference module realizes the spatiotemporal change mapping of pollutant coarse-grained distribution to fine-grained distribution based on the time-related characteristics of pollutant diffusion, specifically including: According to the specified upsampling factor r, the time-dependent characteristics f of the pollutant diffusion are output Perform upsampling(·) to obtain the inferred fine-grained pollution distribution map Through the mean square error loss function Optimizing the inference model: T represents the total number of moments in the grid sequence of the fine-grained pollution distribution map, represents the pollutant value of the fine-grained pollution distribution grid at a specific location at the t-th moment; Y(t) represents the pollutant value of the actual fine-grained pollution distribution grid at a specific location at the t-th moment.
6. 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.
7. 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.