Three-dimensional occupied grid network sensing method and device and medium

Abstract

The invention relates to a three-dimensional occupied grid network sensing method, which comprises the following steps of: acquiring a multi-view camera image shot by automatic driving of a vehicle, inputting the multi-view camera image into an image processing backbone network, and extracting two-dimensional image features; performing view angle conversion on the two-dimensional image features to obtain three-dimensional body features; constructing matching cost body features according to the geometric correspondence of the sampling points at different moments; a convolutional neural network is adopted to extract weights containing depth correction information from the matching cost body features, the weights act on the three-dimensional body features at the current moment, and the three-dimensional body features after depth information correction are obtained; and inputting the three-dimensional body features into an occupied grid decoder, projecting the three-dimensional body features to a discrete semantic space, and generating a three-dimensional semantic occupied grid. Compared with the prior art, by constructing the matching cost body, more accurate depth information and geometric understanding can be obtained, the calculation overhead under a long time sequence is reduced, and accurate and efficient modeling of the automatic driving environment can be realized.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving, and in particular to a three-dimensional occupancy grid network perception method, device and medium. Background Art

[0002] Three-dimensional occupancy grid network perception is an important research direction in the field of autonomous driving. It improves the environmental perception capability of the autonomous driving system by predicting occupancy in three-dimensional space.

[0003] Three-dimensional occupancy grid network perception is a perception algorithm that divides the three-dimensional space into multiple small cubes (voxels). Each voxel represents a unit in the space. Perception is achieved by predicting whether these voxels are occupied and the category information to which the occupied voxels belong. This enables a more accurate description of the geometric shape and position of objects, thereby improving the perception accuracy of the autonomous driving system.

[0004] The existing 3D occupancy grid network perception method based on temporal fusion has the following defects:

[0005] 1) Methods based on coordinate system alignment include aligning the coordinate systems of the bird's-eye view feature representations at different times, using the self-attention mechanism for feature fusion, and concatenating features after aligning the coordinate systems, inputting them into the convolutional neural network, and performing temporal information fusion. However, the above methods only implicitly utilize the geometric correspondence and lack a sufficiently robust understanding of three-dimensional spatial information, which greatly limits the accuracy of autonomous driving environment modeling.

[0006] 2) Stereo matching method: Construct matching cost volume features from images captured at different times in the time series, and use geometric constraints to obtain more accurate depth information. However, for multi-camera pure vision tasks in autonomous driving, the geometric growth of the number of image pairs may lead to huge computational overhead when fusing longer time spans.

[0007] Therefore, there is an urgent need to design a three-dimensional occupancy grid network perception method that can achieve more accurate and efficient modeling of the environment surrounding autonomous driving. Summary of the invention

[0008] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide an autonomous driving three-dimensional occupancy grid network perception method that can achieve more accurate and efficient modeling of the autonomous driving surrounding environment.

[0009] The purpose of the present invention can be achieved by the following technical solutions:

[0010] According to a first aspect of the present invention, there is provided a three-dimensional occupancy grid network perception method, comprising:

[0011] Collect multi-camera images taken by the vehicle's automatic driving, input the multi-camera images into an image processing backbone network, and extract two-dimensional image features of multiple specifications;

[0012] Performing perspective conversion on the two-dimensional image features to obtain three-dimensional body features;

[0013] Constructing a matching cost according to the geometric correspondence of the sampling points at different times, and extracting a matching cost volume feature from the three-dimensional volume feature;

[0014] A convolutional neural network is used to extract weights containing depth correction information from the matching cost volume features, and the weights are applied to the three-dimensional volume features at the current moment to obtain the three-dimensional volume features after depth information correction;

[0015] The 3D volume features corrected by depth information are input into the occupancy grid decoder, and the 3D volume features are projected into a discrete semantic space to generate the final 3D semantic occupancy grid.

[0016] Preferably, the multi-camera images taken by the vehicle's automatic driving are collected, and the multi-camera images are input into an image processing backbone network to extract two-dimensional image features of multiple specifications, specifically:

[0017] Collect multi-camera images taken by the vehicle's autonomous driving in, represents the image taken by the i-th camera at time T, and L is the total number of cameras;

[0018] The multi-camera image Input into the image processing backbone network to extract multi-specification two-dimensional image features

[0019] Preferably, the two-dimensional image features are transformed from a perspective to obtain three-dimensional features, specifically: the BEVFormer model is adopted, and the learnable three-dimensional query points are used to transform the two-dimensional image features from a perspective through a cross-attention mechanism to obtain three-dimensional features, and the expression is:

[0020]

[0021] Where: Q is the learnable 3D query point in the cross attention mechanism; L is the total number of cameras, M is the total number of reference points; P is the projection function, P(i,j) represents the 2D coordinate position obtained by projecting the jth reference point into the image coordinate system of the i-th camera; J is the 2D image feature, DeformAttn represents the deformable attention mechanism; H, W, Z, C are the X, Y, Z and feature dimensions of the 3D volume feature, respectively.

[0022] Preferably, a BEVDet-like network model is used to perform perspective conversion on the two-dimensional image features to obtain three-dimensional body features, specifically: the two-dimensional image features are input into a depth prediction network to predict a depth classification score; using the predicted depth classification score, a perspective converter is used to project the two-dimensional image features into a predefined three-dimensional point cloud framework, and BEV features are generated by applying a pooling operation along the vertical direction to obtain three-dimensional body features.

[0023] Preferably, the step of constructing a matching cost according to the geometric correspondence between the sampling points at different times and extracting a matching cost volume feature from the three-dimensional volume feature includes:

[0024] Input the 3D volume features {V T ,V T-1 ,…,V T-K+1}, where K represents the total number of frames of input data;

[0025] The three-dimensional feature V at the current time T T For each voxel point, the three-dimensional features of N points are sampled at a fixed interval on its line of sight, denoted as Among them, the position of the vehicle is taken as the center of the three-dimensional coordinate system, and the line connecting the center of the three-dimensional coordinate system and the corresponding voxel point is taken as the line of sight;

[0026] Use the projection matrix to align the three-dimensional coordinate system of all historical moments with the current moment T;

[0027] Project all sampling points at the current moment T to the coordinate system of the historical moment, and sample the three-dimensional features of the corresponding positions at the historical moment, recorded as

[0028] The 3D volume features at all times are concatenated in the feature dimension to obtain the matching cost volume feature F∈R H ×W×Z×(K×N)×C , H, W, Z, and C are the sizes of the X, Y, Z, and feature dimensions of the three-dimensional feature respectively.

[0029] Preferably, the convolutional neural network is used to extract weights containing depth correction information from the matching cost volume features, and the weights are applied to the three-dimensional volume features at the current moment to obtain the three-dimensional volume features after depth information correction, specifically:

[0030] Input the matching cost volume features into the convolutional neural network to obtain the weight W∈R used to assist in correcting the depth information H×W×Z , H, W, Z are X, Y, Z of the three-dimensional features respectively, and the normalized weights are supervised by using the binary cross entropy loss function;

[0031] The weight after supervised learning is element-wise multiplied with the three-dimensional volume features at the current moment to obtain the three-dimensional volume features corrected by the depth information.

[0032] Preferably, the binary cross entropy loss function is expressed as:

[0033]

[0034] Where: w j is the jth element in weight W; Indicates that the jth voxel in the three-dimensional occupancy grid truth value is not occupied, Indicates that the jth voxel in the three-dimensional occupancy grid truth is occupied by one of the categories.

[0035] Preferably, the modified three-dimensional volume features are used as input of an occupancy grid decoder to generate a final three-dimensional semantic occupancy grid, specifically:

[0036] The corrected three-dimensional feature V T As the input of the occupancy grid decoder, the 3D volume features are projected into a discrete semantic space through deconvolution to generate the final 3D semantic occupancy grid Y∈{c 0 ,c 1 ,…,c M} G×W×Z , c i is the label of the i-th category, and M is the total number of categories.

[0037] According to a second aspect of the present invention, there is provided an electronic device, comprising a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the program, any one of the methods described above is implemented.

[0038] According to a third aspect of the present invention, there is provided a computer-readable storage medium having a computer program stored thereon, wherein when the program is executed by a processor, any one of the methods described above is implemented.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] (1) By constructing matching proxy features in a three-dimensional feature space, the present invention can explicitly model the geometric correspondence at different times, compared with the method based on coordinate system alignment, and obtain more accurate prediction results of the surrounding environment perception of autonomous driving.

[0041] (2) The present invention utilizes the fact that multiple sampling points on the same line of sight at the current moment will not have parallax on the same line of sight after being projected to the historical moment, thereby providing additional geometric information, which can reduce the uncertainty of depth information in the autonomous driving environment image. Compared with the existing method of constructing matching cost volume features from image pairs at different moments, the present invention has a smaller computational overhead.

[0042] (3) The BEVFormer model is used to extract three-dimensional features from two-dimensional image features using learnable three-dimensional query points through a cross-attention mechanism. Three-dimensional features can be directly constructed, bypassing the explicit depth estimation step, thus avoiding further amplification of the uncertainty error of depth information in the autonomous driving environment image and improving the accuracy of subsequent perception predictions.

[0043] (4) The BEVDet-type network model is used to transform the perspective of two-dimensional image features to obtain three-dimensional features. The depth prediction network, perspective converter and pooling operator in it can be freely adjusted and set according to the actual autonomous driving environment, which is more flexible. The depth information obtained in the intermediate steps can be used as an additional optimization target for supervised learning, which can improve the subsequent perception prediction effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 is a flow chart of the method of the present invention;

[0045] Figure 2 It is a schematic diagram of the method architecture of the present invention. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0047] Example

[0048] like Figure 1 and Figure 2 As shown, this embodiment provides a three-dimensional occupancy grid network perception method for three-dimensional modeling of the vehicle surrounding environment during autonomous driving, and the specific implementation includes the following steps:

[0049] S1. Collect multi-camera images taken by the vehicle's autonomous driving, input the multi-camera images into the image processing backbone network, and extract multi-specification two-dimensional image features, specifically:

[0050] The input data in this embodiment are images taken by multiple cameras during the vehicle's automatic driving process, that is, images taken by multiple cameras at different angles on the vehicle during the vehicle's driving, and these input photos are continuous in time. The input image of the camera numbered i at time T is denoted as The temporal fusion in this embodiment uses the input images of the current moment and the historical moment at the same time, denoted as L is the total number of cameras.

[0051] Multi-camera images Input into the image processing backbone network to extract multi-specification two-dimensional image features in, yes Multi-scale 2D features corresponding to the image.

[0052] S2. Perform perspective transformation on the two-dimensional image features to obtain three-dimensional features, specifically:

[0053] In this embodiment, the BEVFormer model is used to transform the perspective of the two-dimensional image features using the learnable three-dimensional query points through the cross-attention mechanism to obtain the three-dimensional body features, which are expressed as:

[0054]

[0055] Where: Q is the learnable 3D query point in the cross attention mechanism; L is the total number of cameras, M is the total number of reference points; P is the projection function, P(i,j) represents the 2D coordinate position obtained by projecting the jth reference point into the image coordinate system of the i-th camera; J is the 2D image feature, DeformAttn represents the deformable attention mechanism; H, W, Z, C are the X, Y, Z and feature dimensions of the 3D volume feature, respectively.

[0056] The BEVFormer model is based on Transformer and variable attention mechanism, and uses the spatial and temporal information of multi-camera images to uniformly represent the surrounding scenes of the autonomous driving vehicle from a bird's-eye view (BEV). Through multi-camera image input, the Transformer's attention mechanism is used to convert the features in the image space into features in the BEV space. In this process, BEVFormer combines the spatial cross-attention and temporal self-attention modules to achieve effective fusion of different camera views and historical BEV feature information. The BEVFormer model framework mainly includes multiple encoder layers, each of which is defined based on the Transformer standard, and adds bird's-eye view query, spatial cross-attention, and temporal self-attention modules.

[0057] The method in BEVFormer uses learnable three-dimensional query points to extract three-dimensional body features from two-dimensional image features through a cross-attention mechanism.

[0058] In this embodiment, the BEVFormer model realizes perspective conversion without explicit depth estimation. In pure visual tasks, the problem of estimating depth from images is uncertain, and the error of this intermediate step will be amplified in the further construction of BEV features, significantly affecting the accuracy of the final result. The BEVFormer model can directly construct three-dimensional features by defining three-dimensional query points in advance and extracting information from two-dimensional image feature points through a cross-attention mechanism, bypassing the step of explicit depth estimation.

[0059] S3, constructing matching costs according to the geometric correspondence of sampling points at different times, obtaining a more accurate spatial understanding, thereby extracting matching cost volume features from three-dimensional volume features. This step is the core of the present invention, such as Figure 2 As shown in the matching cost body timing module in, specifically:

[0060] 1) Input the 3D volume features {V T ,V T-1 ,…,V T-K+1};

[0061] 2) The three-dimensional feature V at the current moment T For each voxel point, the three-dimensional features of N points are sampled at a fixed interval on its line of sight, denoted as In general, the position of the vehicle is taken as the center of the three-dimensional coordinate system, and the line connecting the center of the three-dimensional coordinate system and the corresponding voxel point is taken as the line of sight;

[0062] 3) Use the projection matrix in the data set to align the 3D coordinate system of all historical moments with the current moment;

[0063] 4) Project all sampling points at the current moment into the coordinate system of the historical moment, and sample the three-dimensional features of the corresponding positions at the historical moment, recorded as

[0064] 5) Concatenate the 3D volume features at all times in the feature dimension to obtain the matching cost volume feature F∈R H ×W×Z×(K×N)×c .

[0065] S4. Use a convolutional neural network to extract weights containing depth correction information from the matching cost volume features, apply the weights to the three-dimensional volume features at the current moment, and obtain the three-dimensional volume features after depth information correction, specifically:

[0066] Input the matching cost volume features into the convolutional neural network to obtain the weight W∈R used to assist in correcting the depth information H×W×z, the binary cross entropy loss function is used to supervise the weights normalized to the interval [0,1], so that the weight W tends to 0 in the unoccupied position and tends to 1 in the occupied position; among them, the binary cross entropy loss function is expressed as:

[0067]

[0068] Where: w j is the jth element in weight W; Indicates that the jth voxel in the three-dimensional occupancy grid truth value is not occupied, Indicates that the jth voxel in the three-dimensional occupancy grid truth is occupied by one of the categories;

[0069] The weight after supervised learning is element-wise multiplied with the three-dimensional volume features at the current moment to obtain the three-dimensional volume features corrected by the depth information.

[0070] S5. Input the three-dimensional volume features after depth information correction into the occupancy grid decoder, project the three-dimensional volume features into a discrete semantic space, and generate a final three-dimensional semantic occupancy grid.

[0071] The corrected three-dimensional feature V T As the input of the occupancy grid decoder, the 3D volume features are projected into a discrete semantic space through deconvolution to generate the final 3D semantic occupancy grid Y∈{c 0 ,c 1 ,…,c M} H×W×Z , c i is the label of the i-th category, and M is the total number of categories.

[0072] The electronic device of the present invention includes a central processing unit (CPU), which can perform various appropriate actions and processes according to computer program instructions stored in a read-only memory (ROM) or loaded from a storage unit into a random access memory (RAM). In the RAM, various programs and data required for device operation can also be stored. The CPU, ROM and RAM are connected to each other via a bus. An input / output (I / O) interface is also connected to the bus.

[0073] Multiple components in the device are connected to the I / O interface, including: input units, such as keyboards, mice, etc.; output units, such as various types of displays, speakers, etc.; storage units, such as disks, optical disks, etc.; and communication units, such as network cards, modems, wireless communication transceivers, etc. The communication unit allows the device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunication networks.

[0074] The processing unit performs the various methods and processes described above, such as methods S1 to S5. For example, in some embodiments, methods S1 to S5 may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program may be loaded and / or installed on the device via a ROM and / or a communication unit. When the computer program is loaded into the RAM and executed by the CPU, one or more steps of methods S1 to S5 described above may be performed. Alternatively, in other embodiments, the CPU may be configured to execute methods S1 to S5 in any other appropriate manner (e.g., by means of firmware).

[0075] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chip systems (SOCs), load programmable logic devices (CPLDs), and the like.

[0076] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code can be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.

[0077] In the context of the present invention, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0078] Example 2

[0079] In this embodiment S2, a BEVDet-type network model is used to perform perspective conversion on two-dimensional image features to obtain three-dimensional body features, specifically: the two-dimensional image features are input into the depth prediction network to predict the depth classification score; the predicted depth classification score is used to project the two-dimensional image features into a predefined three-dimensional point cloud framework using a perspective converter, and the BEV features are generated by applying a pooling operation in the vertical direction to obtain three-dimensional body features. Among them, the depth prediction network, perspective converter and pooling operator can freely design different network architectures, such as mainstream convolutional neural networks or attention mechanisms. If it is not limited to pure visual tasks, and there is depth information or point cloud information in the data set, the depth information and three-dimensional pseudo point cloud obtained in the intermediate step can be used as additional optimization targets for supervised learning, which can improve the prediction effect.

[0080] The other settings in this embodiment are the same as those in Embodiment 1.

[0081] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of various equivalent modifications or substitutions within the technical scope disclosed by the present invention, and these modifications or substitutions should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A three-dimensional occupancy grid network perception method, characterized in that: include: Collect multi-camera images taken by the vehicle's automatic driving, input the multi-camera images into an image processing backbone network, and extract two-dimensional image features of multiple specifications; Performing perspective conversion on the two-dimensional image features to obtain three-dimensional body features; Constructing a matching cost according to the geometric correspondence of the sampling points at different times, and extracting a matching cost volume feature from the three-dimensional volume feature; A convolutional neural network is used to extract weights containing depth correction information from the matching cost volume features, and the weights are applied to the three-dimensional volume features at the current moment to obtain the three-dimensional volume features after depth information correction; The 3D volume features corrected by depth information are input into the occupancy grid decoder, and the 3D volume features are projected into a discrete semantic space to generate the final 3D semantic occupancy grid.

2. A three-dimensional occupancy grid network perception method according to claim 1, characterized in that: The multi-camera images taken by the automatic driving of the vehicle are collected, and the multi-camera images are input into the image processing backbone network to extract two-dimensional image features of multiple specifications, specifically: Collect multi-camera images taken by the vehicle's autonomous driving in, represents the image taken by the i-th camera at time T, and L is the total number of cameras; The multi-camera image Input into the image processing backbone network to extract multi-specification two-dimensional image features 3. A three-dimensional occupancy grid network perception method according to claim 1, characterized in that: The perspective transformation of the two-dimensional image features is performed to obtain three-dimensional features. Specifically, the BEVFormer model is used to transform the perspective of the two-dimensional image features through the cross-attention mechanism using learnable three-dimensional query points to obtain three-dimensional features. The expression is: Where: Q is the learnable 3D query point in the cross attention mechanism; L is the total number of cameras, M is the total number of reference points; P is the projection function, P(i,j) represents the 2D coordinate position obtained by projecting the jth reference point into the image coordinate system of the i-th camera; J is the 2D image feature, DeformAttn represents the deformable attention mechanism; H, W, Z, C are the X, Y, Z and feature dimensions of the 3D volume feature, respectively.

4. A three-dimensional occupancy grid network perception method according to claim 1, characterized in that: A BEVDet-like network model is used to perform perspective conversion on the two-dimensional image features to obtain three-dimensional body features. Specifically, the two-dimensional image features are input into a depth prediction network to predict a depth classification score. Using the predicted depth classification score, a perspective converter is used to project the two-dimensional image features into a predefined three-dimensional point cloud framework, and a pooling operation is applied along the vertical direction to generate BEV features to obtain three-dimensional body features.

5. A three-dimensional occupancy grid network perception method according to claim 1, characterized in that: The matching cost is constructed according to the geometric correspondence between the sampling points at different times, and the matching cost volume features are extracted from the three-dimensional volume features, including: Input the 3D volume features {V T ,V T-1 ,…,V T-K+1 }, where K represents the total number of frames of input data; The three-dimensional feature V at the current time T T For each voxel point, the three-dimensional features of N points are sampled at a fixed interval on its line of sight, denoted as Among them, the position of the vehicle is taken as the center of the three-dimensional coordinate system, and the line connecting the center of the three-dimensional coordinate system and the corresponding voxel point is taken as the line of sight; Use the projection matrix to align the three-dimensional coordinate system of all historical moments with the current moment T; Project all sampling points at the current moment T to the coordinate system of the historical moment, and sample the three-dimensional features of the corresponding positions at the historical moment, recorded as The 3D volume features at all times are concatenated in the feature dimension to obtain the matching cost volume feature F∈R H×W×Z×(K×N)×C , H, W, Z, and C are the sizes of the X, Y, Z, and feature dimensions of the three-dimensional feature respectively.

6. A three-dimensional occupancy grid network perception method according to claim 1, characterized in that: The convolutional neural network is used to extract the weight containing the depth correction information from the matching cost volume feature, and the weight is applied to the three-dimensional volume feature at the current moment to obtain the three-dimensional volume feature after the depth information is corrected, specifically: Input the matching cost volume features into the convolutional neural network to obtain the weight W∈R used to assist in correcting the depth information H ×W×Z , H, W, Z are X, Y, Z of the three-dimensional features respectively, and the normalized weights are supervised by using the binary cross entropy loss function; The weight after supervised learning is element-wise multiplied with the three-dimensional volume features at the current moment to obtain the three-dimensional volume features corrected by the depth information.

7. A three-dimensional occupancy grid network perception method according to claim 6, characterized in that: The binary cross entropy loss function is expressed as: Where: w j is the jth element in weight W; Indicates that the jth voxel in the three-dimensional occupancy grid truth value is not occupied, Indicates that the jth voxel in the three-dimensional occupancy grid truth is occupied by one of the categories.

8. A three-dimensional occupancy grid network perception method according to claim 1, characterized in that: The modified 3D volume features are used as the input of the occupancy grid decoder to generate the final 3D semantic occupancy grid, specifically: The corrected three-dimensional feature V T As the input of the occupancy grid decoder, the 3D volume features are projected into a discrete semantic space through deconvolution to generate the final 3D semantic occupancy grid Y∈{c0,c1,…,c M } H×W×Z , c i is the label of the i-th category, and M is the total number of categories.

9. An electronic device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the program, the method according to any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images