Multi-source video collaborative imaging method and device

Abstract

The present invention discloses a multi-source video collaborative imaging method and device, belonging to the field of image processing technology. The method extracts features from different source videos separately to obtain a feature pyramid; constructs a geometric modeling subnetwork and a radiation modeling subnetwork to obtain the density field and radiation field of the source video; performs mutation detection based on the density field of any source video to obtain the occluded area of ​​the source video; performs color complementation on the occluded area based on other source videos to obtain a complemented radiation field; uses the complemented radiation field to replace the occluded area in the radiation field to obtain a scene field; and fuses the scene fields of all source videos to synthesize an image. The method and device disclosed in the present invention ultimately obtain an image with high clarity and good picture coherence.

Description

Technical Field

[0001] The present invention relates to a multi-source video collaborative imaging method and device, belonging to the technical field of image processing. Background Art

[0002] In many video application scenarios, such as live broadcasts of large-scale events, intelligent driving, and complex industrial monitoring, there are generally multiple video source inputs, and multiple video sources need to be fused to obtain a comprehensive video.

[0003] Traditional multi-video processing methods are relatively simple, typically simply splicing or switching multiple video frames for display, failing to fully tap into the information value of multiple video sources. This processing method can result in poor image coherence and significant loss of detail in the final image, making it difficult to meet the requirements of scenarios with high image quality and information integrity.

[0004] Especially in actual use, due to the differences in device models, positions, angles, and installation methods, different video source shooting devices may have various anomalies in the videos obtained, including jitter, partial obstruction of the field of view, partial or full area exposure, color deviation, etc.

[0005] In addition, in order to obtain a wider field of view, existing shooting equipment, especially some vehicle-mounted lenses and some outdoor live broadcast lenses, will use wide-angle lenses, which will produce distortion and other phenomena, resulting in serious deviations in the positions of objects in the fused video. If this deviation is applied to intelligent driving, it may lead to wrong judgments during intelligent driving and even cause serious consequences; if applied to vehicle-mounted camera live broadcasts, it will cause picture shaking and distortion, affecting the viewing effect.

[0006] Therefore, it is necessary to conduct more in-depth research on the existing multi-source video collaborative imaging methods to solve the above problems. Summary of the Invention

[0007] In order to overcome the above problems, in-depth research was conducted and a multi-source video collaborative imaging method was proposed, which includes the following steps:

[0008] S1. Obtain motion trajectories of different cameras and preprocess all video sources separately, wherein the preprocessing includes dynamic stabilization and distortion correction;

[0009] S2, based on the pre-processed video and the motion trajectory of each camera, perform spatial and temporal alignment between different source videos;

[0010] S3, extract features from different videos after spatial and temporal alignment to obtain feature pyramids;

[0011] S4. Construct a geometric modeling subnetwork to obtain the density field of any source video based on the feature pyramid of the source video to characterize the geometric structure of the scene;

[0012] Construct a radiation modeling subnetwork to obtain the radiation field of any source video based on the density field of the source video, which is used to characterize the surface optical properties of objects in the scene;

[0013] Construct a dynamic field modeling sub-network to obtain the dynamic field of any source video based on the radiation field of the source video, which is used to characterize the dynamic changes of the scene;

[0014] S5. Perform mutation detection based on the density field of any source video to obtain the occluded area of ​​the source video, and perform color complementation on the occluded area based on other source videos to obtain a complemented radiation field;

[0015] The scene field is obtained by replacing the blocked area in the radiation field with the completed radiation field;

[0016] S6. Fusing the scenes of all source videos to synthesize an image.

[0017] In a preferred embodiment, in S2, during the timing alignment process, virtual frames are generated so that the frame rates of all video sources are the same, including the following steps:

[0018] For any source video,

[0019] S221, obtaining the optical flow field of the video;

[0020] S222. Based on the motion trajectory of its camera, predict the relative motion between adjacent frames of the video and obtain a global motion optical flow field prediction;

[0021] S223, aligning the optical flow field with the global motion optical flow field prediction to generate residual optical flow;

[0022] S224: Generate a virtual frame based on the residual optical flow and add it to the actual frame, so that different source videos can achieve frame alignment.

[0023] In a preferred embodiment, in S4, the geometric modeling subnetwork is a multi-layer MLP structure, which takes the feature pyramid and the corresponding timestamp as input and outputs the density field at different times.

[0024] In a preferred embodiment, the radiation modeling sub-network is a multi-layer MLP structure, and its input is the density field and the viewing direction of the camera lens.

[0025] In a preferred embodiment, the dynamic field modeling subnetwork includes a gated recurrent unit and a fully connected network, and its inputs are density field and radiation field.

[0026] In a preferred embodiment, a distortion field sub-network is also constructed in S4 to perform distortion correction on the radiation field. The distortion field sub-network is a fully connected network with a multi-layer MLP structure. Its input is normalized pixel coordinates and camera focal length parameters, and its output is the offset of the pixel coordinates. The pixel coordinate offset is used to correct the pixel coordinates in the radiation modeling sub-network.

[0027] In a preferred embodiment, in S5, the completion includes the following sub-steps:

[0028] S51, extracting feature blocks from feature pyramids corresponding to other source videos according to the position of the occlusion area of ​​the current video source;

[0029] S52, projecting the extracted feature block to the current video source viewing angle coordinate system to obtain a projected feature block;

[0030] S53, obtaining similarities between the occluded area and the projection feature blocks corresponding to different source videos;

[0031] S54, using similarity as weight, fusing multiple projected feature blocks to obtain a completed feature pyramid;

[0032] S55. Based on the completed feature pyramid, the completed radiation field is obtained through the geometric modeling sub-network and the radiation modeling sub-network.

[0033] In a preferred embodiment, in S6, different scene fields in the fusion process have different weights, and the weights are obtained based on the viewing angle of the camera shooting the source video.

[0034] The present invention also provides an electronic device, comprising:

[0035] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform any one of the above methods.

[0036] The present invention also provides a computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable the computer to execute any one of the above methods.

[0037] The beneficial effects of the present invention include:

[0038] (1) The final image has high clarity and good picture continuity;

[0039] (2) Strong anti-shake capability, solving the problem of image distortion caused by field of view obstruction;

[0040] (3) Solved the problem of object position deviation caused by wide-angle lens. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 The figure is a flowchart of a multi-source video collaborative imaging method according to a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0042] The present invention will be described in further detail below with reference to the accompanying drawings and examples, through which the features and advantages of the present invention will become more clearly understood.

[0043] The word "exemplary" is used exclusively herein to mean "serving as an example, example, or illustration." Any embodiment described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments. Although various aspects of the embodiments are shown in the drawings, the drawings are not necessarily drawn to scale unless otherwise indicated.

[0044] According to a multi-source video collaborative imaging method provided by the present invention, Figure 1 As shown, the following steps are included:

[0045] S1. Obtain motion trajectories of different cameras and preprocess all video sources separately, wherein the preprocessing includes dynamic stabilization and distortion correction;

[0046] S2, based on the pre-processed video and the motion trajectory of each camera, perform spatial and temporal alignment between different source videos;

[0047] S3, extract features from different videos after spatial and temporal alignment to obtain feature pyramids;

[0048] S4. Construct a geometric modeling subnetwork to obtain the density field of any source video based on the feature pyramid of the source video to characterize the geometric structure of the scene;

[0049] Construct a radiation modeling subnetwork to obtain the radiation field of any source video based on the density field of the source video, which is used to characterize the surface optical properties of objects in the scene;

[0050] Construct a dynamic field modeling sub-network to obtain the dynamic field of any source video based on the radiation field of the source video, which is used to characterize the dynamic changes of the scene;

[0051] S5. Perform mutation detection based on the density field of any source video to obtain the occluded area of ​​the source video, and perform color complementation on the occluded area based on other source videos to obtain a complemented radiation field;

[0052] The scene field is obtained by replacing the blocked area in the radiation field with the completed radiation field;

[0053] S6. Fusing the scenes of all source videos to synthesize an image.

[0054] In S1, the motion trajectory of the camera is the camera 6DOF parameters and timestamp, which are obtained based on the camera's IMU data. The 6DOF parameters are commonly used parameters in computer vision. The present invention does not limit the specific method of obtaining them, and those skilled in the art can freely perform it based on experience.

[0055] The dynamic stabilization is used to eliminate involuntary motion jitter. The optical flow method is used to calculate the local motion of adjacent frames, which is fused with the IMU data. The current frame is reversely warped through the motion compensation matrix to eliminate high-frequency jitter.

[0056] The distortion correction is used to reduce the effect of lens distortion and apply inverse distortion to perform pixel remapping based on camera calibration parameters (focal length, distortion coefficient).

[0057] Preferably, in S1, the preprocessing further includes exposure balancing processing, by constructing a brightness histogram pyramid, aligning the dynamic range of each camera, and then adjusting the overexposed / underexposed areas through an adaptive gamma curve.

[0058] In S2, different source videos are spatially aligned by building geometric relationships between different cameras.

[0059] Preferably, spatial alignment is performed by the following steps:

[0060] S211, performing ORB feature extraction on each pre-processed video to obtain feature points;

[0061] S212, matching feature points of different source videos to obtain multiple matching point pairs;

[0062] S213 : Perform linear triangulation on the matching point pairs of each source video to generate a 3D point cloud of the source video.

[0063] In S211, ORB feature extraction is a commonly used method in SLAM, and its specific process is not described in detail in the present invention.

[0064] In S212 , a distance measurement method is used to obtain the similarity between the feature points, and a matching result is obtained based on the similarity. For example, the similarity between the feature points is expressed by the Hamming distance or the Euclidean distance, thereby achieving matching.

[0065] In S213, linear triangulation is a commonly used method in SLAM and will not be described in detail in the present invention.

[0066] Preferably, there is also step S214 of optimizing the 3D point cloud through bundle adjustment. Bundle adjustment (BA) is a method for optimizing multi-view geometric parameters by minimizing reprojection errors. It is widely used in photogrammetry and computer vision, and its specific process is not described in detail in this invention.

[0067] The timing alignment refers to the timing alignment of different source videos. There will be a time deviation between the original frame sequences shot by different cameras, that is, the shooting frame rates are different. This deviation will cause the subsequent composite image to be blurred and blurry.

[0068] Furthermore, by generating virtual frames, the frame rates of all video sources are made the same.

[0069] Preferably, the timing alignment comprises the following steps:

[0070] For any source video,

[0071] S221, obtaining the optical flow field of the video;

[0072] S222: Based on the motion trajectory of the camera, predict the relative motion between adjacent frames of the video and obtain a global motion optical flow field prediction;

[0073] S223, aligning the optical flow field with the global motion optical flow field prediction to generate residual optical flow;

[0074] S224: Generate a virtual frame based on the residual optical flow and add it to the actual frame, so that different source videos can achieve frame alignment.

[0075] In S221, the optical flow field is obtained based on the Horn-Schunck method.

[0076] In S222, the global motion prediction is expressed as:

[0077]

[0078] in, express t The position of the camera at a given moment is obtained from the camera's motion trajectory. express t Always t Motion prediction at time +1, that is, the relative motion between adjacent frames of the video.

[0079] The global motion optical flow field prediction Expressed as:

[0080]

[0081] in, is the projection function, is the pixel coordinate, For depth.

[0082] In S223, the residual optical flow is the difference between the optical flow field and the global motion optical flow field prediction, and represents the real motion of the local dynamic object.

[0083] In S224, the virtual synchronization frame is represented as:

[0084]

[0085] Among them, (x, y) is the pixel coordinate, is the virtual synchronization frame corresponding to the virtual time τ, is the set of moments of the virtual frame, is the time-integrated offset of the residual perfusion, is the weight coefficient, is the real frame at time t.

[0086] Wherein, the weight coefficient is set as:

[0087]

[0088] in, is the moment in the virtual frame, represents the residual optical flow, The weight decay rate can be set.

[0089] In S3, the feature extraction is implemented using a ConvNeXt network. The ConvNeXt network is a supervised convolutional neural network that is widely used in different visual processing. Its structure is not described in detail in the present invention.

[0090] Preferably, the feature pyramid includes feature tensors of at least 4 scales.

[0091] In S4, the geometric modeling subnetwork is a multi-layer MLP (multi-layer perceptron network) structure, which takes the feature pyramid and the corresponding timestamp as input and outputs the density field at different times.

[0092] By adding the input of the time dimension, the geometric field is made to have time-varying characteristics to adapt to the subsequent fusion process.

[0093] In a preferred embodiment, the input of the geometric modeling sub-network also includes the depth of the image to more accurately characterize the geometric structure of the scene.

[0094] Preferably, the geometric modeling subnetwork is a 5-layer MLP structure, the first four layers use the LeakyReLU activation function, and the last layer uses the Softplus activation function.

[0095] The radiation field is the basic color of the image frame pixels, and the optical properties of the objects in the image are represented by the basic color.

[0096] Preferably, the input of the radiation modeling subnetwork also includes the viewing direction of the camera lens. More preferably, the viewing direction is input in the form of spherical harmonic encoding, for example, the viewing direction is expanded into a 9-dimensional vector by a third-order spherical harmonic function.

[0097] According to the present invention, the viewing direction is spliced ​​onto the density field and used together as the input of the radiation modeling sub-network.

[0098] By increasing the viewing angle direction and introducing dynamic features in the prediction process, the accuracy of color prediction corresponding to different source videos is improved.

[0099] According to the present invention, the radiation modeling subnetwork is a multi-layer MLP structure.

[0100] Preferably, the radiation modeling subnetwork includes a 3-layer MLP structure, the first two layers use ReLU activation functions, and the last layer uses Sigmoid function.

[0101] The dynamic field is the rate of change of color over time, reflecting the motion blur effect.

[0102] Preferably, the inputs of the dynamic field modeling subnetwork are density field and radiation field.

[0103] Preferably, the dynamic field modeling subnetwork includes a gated recurrent unit (GRU) and a fully connected network. The density field and the radiation field are spliced ​​and input into the gated recurrent unit. The output of the gated recurrent unit passes through a multi-layer fully connected network to output the dynamic field.

[0104] In the present invention, the density field is used as one of the inputs to play the role of a carrier signal, so that the dynamic field modeling subnetwork can selectively absorb the characteristic information of the radiation field and improve the accuracy.

[0105] In the present invention, by setting up a geometric modeling subnetwork, a radiation modeling subnetwork and a dynamic field modeling subnetwork, different physical effects are separated, the parameter interpretability is improved, the accuracy of the model is improved, and the blurring effect caused by motion is greatly reduced.

[0106] Preferably, in S4, a distortion field sub-network is also constructed to perform distortion correction on the radiation field.

[0107] The distortion field sub-network is a fully connected network with a multi-layer MLP structure, preferably a 4-layer MLP structure. The ReLU activation function is used between the multi-layer MLP structures. Its input is the normalized pixel coordinates and camera focal length parameters, and the output is the offset of the pixel coordinates.

[0108] Preferably, the output layer of the distortion field sub-network uses a Tanh function to constrain the offset range to prevent overcorrection.

[0109] Furthermore, the pixel coordinate offset is used to correct the pixel coordinates in the radiation modeling subnetwork, thereby participating in the prediction of the radiation field and achieving the effect of secondary correction of distortion.

[0110] According to the present invention, the distortion field sub-network splices the coordinates and focal length parameters of each pixel, predicts the coordinate offset of the position through the MLP layer, and then dynamically compensates for the distortion.

[0111] Furthermore, the loss of the distortion field sub-network is set to the pixel difference between the synthesized image and the original uncorrected frame, so that the distortion field sub-network can learn the deformation of the real distortion.

[0112] According to the present invention, by splicing the coordinates and focal length parameters of each pixel, the consistency of optical characteristics of different cameras is maintained through the input of the focal length parameters, and the compensation strategies of different cameras are automatically adapted.

[0113] Furthermore, in the training stage of the present invention, it is necessary to freeze the distorted field sub-network first, prioritize the optimization of the geometric modeling sub-network, the radiation modeling sub-network and the dynamic field modeling sub-network, and then unfreeze the distorted field sub-network after the optimization is completed, and then train all networks to obtain the final network parameters.

[0114] Traditional distortion correction is performed through camera calibration parameters, such as the preprocessing process in this invention. However, research has found that this method is not very effective during outdoor dynamic lens shooting. Analysis shows that distortion correction in the preprocessing process can only eliminate the main geometric distortion of the lens (such as barrel distortion and pincushion distortion), but has no effect on dynamic distortion such as temperature drift, water vapor refraction, and thermal air distortion, which are common phenomena in automobiles. In this invention, by setting up a distortion field sub-network, a secondary correction of distortion is performed, which can more accurately compensate for distortion, thereby significantly improving the jitter and blurred boundaries of the synthesized image area.

[0115] In S5, the depth change rate in the density field is calculated. When it is higher than a threshold, it is considered that a mutation occurs, and the mutation area is the occlusion area.

[0116] In the present invention, the specific value of the threshold is not limited, and those skilled in the art can freely set it according to actual conditions.

[0117] According to the present invention, the complemented radiation field is a color distribution field in a three-dimensional space. Preferably, the complementation includes the following sub-steps:

[0118] S51, extracting feature blocks from feature pyramids corresponding to other source videos according to the position of the occlusion area of ​​the current video source;

[0119] S52, projecting the extracted feature block to the current video source viewing angle coordinate system to obtain a projected feature block;

[0120] S53, obtaining similarities between the occluded area and the projection feature blocks corresponding to different source videos;

[0121] S54, using similarity as weight, fusing multiple projected feature blocks to obtain a completed feature pyramid;

[0122] S55. Based on the completed feature pyramid, the completed radiation field is obtained through the geometric modeling sub-network and the radiation modeling sub-network.

[0123] Preferably, the method further includes S56 , smoothing the boundary of the supplemented radiation field so that the boundary is smoothly connected to the non-blocked area of ​​the current video source.

[0124] In S51, since there are multiple source videos, there are also multiple feature blocks to be extracted.

[0125] Preferably, before extracting feature blocks from feature pyramids of other source videos, the source video is also detected to confirm whether it is also blocked. If blocked, feature blocks of the video are not extracted.

[0126] In S52, since different source videos are shot by cameras with different positions and angles, the feature blocks are projected onto the current view coordinate system according to the camera's motion trajectory to resolve the spatial dislocation caused by the view difference.

[0127] In S53, the similarity is preferably calculated using cosine similarity. Specifically, the cosine similarity of the features at each position in the occluded area and the corresponding position of the feature projection feature block is calculated, and the set of similarities at different positions is used as the weight matrix of the video source.

[0128] Preferably, the weight matrix also includes the dynamic field of other video sources at that location, i.e., the product of similarity and the dynamic field is used as the weight matrix for that video source. The introduction of the dynamic field suppresses time-varying gradients in the complement area, avoiding non-physical motion in the repaired area.

[0129] In S54, the projected feature blocks of different source videos are fused according to the weight matrix to obtain a completed feature pyramid.

[0130] In S56 , the dynamic field corresponding to the boundary position of the occluded area of ​​the current video source is used as the confidence level to smooth the boundary of the supplemented radiation field.

[0131] The completion method in this invention integrates the color, temporal coherence and motion trend of different source videos, making the completed image more accurate and significantly improving the PSNR of dynamic area completion, especially in scenarios such as water vapor occlusion, reflection occlusion, and smoke occlusion.

[0132] In S6, preferably, different scene fields in the fusion process have different weights, and the weights are obtained based on the viewing angle of the camera shooting the source video.

[0133] Preferably, the weight is expressed as:

[0134]

[0135] in, Indicates the The weight of a video source at point (x, y), is a hyperparameter; Indicates the Shooting direction of the video source With Fusion Rendering Angle.

[0136] Hyperparameters The settings control the intensity of the viewing angle attenuation and determine the speed at which the weight decreases with the viewing angle difference, ensuring a smooth transition between adjacent viewing angles and eliminating seams.

[0137] Obviously, the weights of different source videos obtained above need to be normalized before use.

[0138] Furthermore, the rendering is performed based on a NeRF (NeRF) method, which will not be described in detail in the present invention.

[0139] Various embodiments of the methods described above in the present invention may be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system comprising at least one programmable processor, which may be a special-purpose or general-purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0140] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in the present disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in the present disclosure can be achieved. This is not limited herein.

Claims

1. A multi-source video collaborative imaging method, characterized in that: The following steps are involved: S1. Obtain motion trajectories of different cameras and preprocess all video sources separately, wherein the preprocessing includes dynamic stabilization and distortion correction; S2, based on the pre-processed video and the motion trajectory of each camera, perform spatial and temporal alignment between different source videos; S3, extract features from different videos after spatial and temporal alignment to obtain feature pyramids; S4. Construct a geometric modeling subnetwork to obtain the density field of any source video based on the feature pyramid of the source video to characterize the geometric structure of the scene; Construct a radiation modeling subnetwork to obtain the radiation field of any source video based on the density field of the source video, which is used to characterize the surface optical properties of objects in the scene; Construct a dynamic field modeling sub-network to obtain the dynamic field of any source video based on the radiation field of the source video, which is used to characterize the dynamic changes of the scene; S5. Perform mutation detection based on the density field of any source video to obtain the occluded area of ​​the source video, and perform color complementation on the occluded area based on other source videos to obtain a complemented radiation field; The scene field is obtained by replacing the blocked area in the radiation field with the completed radiation field; S6. Fusing the scenes of all source videos to synthesize an image.

2. The multi-source video collaborative imaging method according to claim 1, characterized in that: In S2, during the timing alignment process, virtual frames are generated to make the frame rates of all video sources the same, including the following steps: For any source video, S221, obtaining the optical flow field of the video; S222: Based on the motion trajectory of the camera, predict the relative motion between adjacent frames of the video and obtain a global motion optical flow field prediction; S223, aligning the optical flow field with the global motion optical flow field prediction to generate residual optical flow; S224: Generate a virtual frame based on the residual optical flow and add it to the actual frame, so that different source videos can achieve frame alignment.

3. The multi-source video collaborative imaging method according to claim 1, characterized in that: In S4, the geometric modeling subnetwork is a multi-layer MLP structure, which takes the feature pyramid and the corresponding timestamp as input and outputs the density field at different times.

4. The multi-source video collaborative imaging method according to claim 1, characterized in that: The radiation modeling sub-network is a multi-layer MLP structure, and its input is the density field and the viewing direction of the camera lens.

5. The multi-source video collaborative imaging method according to claim 1, characterized in that: The dynamic field modeling subnetwork includes a gated recurrent unit and a fully connected network, and its inputs are density field and radiation field.

6. The multi-source video collaborative imaging method according to claim 1, characterized in that: S4 also constructs a distortion field sub-network to perform distortion correction on the radiation field. The distortion field sub-network is a fully connected network with a multi-layer MLP structure. Its input is normalized pixel coordinates and camera focal length parameters, and its output is the offset of the pixel coordinates. The pixel coordinate offset is used to correct the pixel coordinates in the radiation modeling sub-network.

7. The multi-source video collaborative imaging method according to claim 1, characterized in that: In S5, the completion includes the following sub-steps: S51, extracting feature blocks from feature pyramids corresponding to other source videos according to the position of the occlusion area of ​​the current video source; S52, projecting the extracted feature block to the current video source viewing angle coordinate system to obtain a projected feature block; S53, obtaining similarities between the occluded area and the projection feature blocks corresponding to different source videos; S54, using similarity as weight, fusing multiple projected feature blocks to obtain a completed feature pyramid; S55. Based on the completed feature pyramid, the completed radiation field is obtained through the geometric modeling sub-network and the radiation modeling sub-network.

8. The multi-source video collaborative imaging method according to claim 1, characterized in that: In S6, different scene fields in the fusion process have different weights, and the weights are obtained based on the viewing angle of the camera shooting the source video.

9. An electronic device, characterized in that: include: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method according to any one of claims 1 to 8.

10. A computer-readable storage medium storing computer instructions, characterized in that: in, The computer instructions are used to cause the computer to execute the method according to any one of claims 1-8.

Images