Scene editing and generating method, system and equipment and storage medium
By extracting and integrating images of the same object from multiple image data, dedistortion processing and depth information recovery, combined with 3DGS training and Gaussian model integration, an editable Gaussian scene model is formed, which solves the problems of low efficiency and insufficient accuracy of traditional scene editing and generation methods, and realizes efficient, accurate and editable scene model construction.
Patent Information
- Application Number
- CN202411761093.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-03
- Publication Date
- 2025-03-11
AI Technical Summary
Traditional scene editing and generation methods rely on NeRF technology or manually creating 3D assets, which have problems such as low efficiency, high computing resources, time-consuming and difficult to ensure scene consistency and authenticity.
By acquiring multiple image data, extracting and integrating images of the same object, obtaining pose information of the object image and performing dedistortion processing, restoring the depth information of the dynamic object image, generating a mesh network, and 3DGS training of the dedistorted image and pose data to build a scene model. At the same time, small object images are trained to obtain a small Gaussian model, integrated into the scene model and associated with the mesh network to form an editable Gaussian scene model.
It improves data processing efficiency, enhances the accuracy of the scene model, and realizes the editability of the scene model, solving the problems of low efficiency and insufficient accuracy of traditional methods.
Smart Images
Figure CN119600207A_ABST
Abstract
Description
[0001] Technology Neighborhood
[0002] The present invention relates to the field of image processing technology, and in particular to a scene editing and generating method, system, device and storage medium. Background Art
[0003] Scene editing and generation are of great significance in the fields of game editing, simulation scene expansion, etc. Traditional methods mostly rely on NeRF technology or manually create 3D assets, but these methods are cumbersome, tedious, and inefficient. Although NeRF technology can project two-dimensional images into three-dimensional space through deep learning, the modeling and rendering process is complex and requires high computing resources. Manually creating 3D assets requires a lot of time and effort, and it is difficult to ensure the consistency and authenticity of the scene. Summary of the invention
[0004] The purpose of the present invention is to address the technical problems existing in the background technology and to propose a scene editing and generating method, system, device and storage medium.
[0005] In order to achieve the above technical objectives, the technical solution adopted by the present invention is as follows:
[0006] A first implementation of the first aspect of the present invention provides a scene editing and generating method, which includes:
[0007] S101, acquiring a plurality of image data, and extracting and integrating a plurality of object images in each of the image data to obtain a plurality of object data sets, wherein the object images in the object data sets are of the same object;
[0008] S102, obtaining position and posture information of a plurality of the object images, and performing dedistortion processing on the plurality of the object images and the corresponding position and posture information to obtain a plurality of dedistorted images;
[0009] S103, restoring and fusing the depth information of each of the dynamic object images, and obtaining a mesh network based on a preset first algorithm;
[0010] S104, performing 3DGS training on the plurality of dedistorted images and the corresponding pose data to obtain a scene model, wherein the dedistorted images include small object images that do not participate in the 3DGS training;
[0011] S105, training a plurality of small object images to obtain a corresponding plurality of small Gaussian models, integrating the plurality of small Gaussian models into the scene model and associating the models with the mesh network to obtain an editable Gaussian scene model.
[0012] Optionally, in a second implementation of the first aspect of the present invention, S101 includes:
[0013] S1011, acquiring a plurality of image data, and using a preset yolov6 algorithm to identify the plurality of image data, to obtain a plurality of object images, wherein the object images are dynamic objects;
[0014] S1012: Integrate the multiple object images belonging to the same object in the multiple image data to obtain multiple object data sets.
[0015] Optionally, in a third implementation of the first aspect of the present invention, S102 includes:
[0016] S1021, using a preset COLMAP tool, processing the plurality of object data sets to obtain an SFM result, wherein the SFM result includes position and posture information of the plurality of object images;
[0017] S1022: Dedistort each of the object images and the corresponding posture information to obtain a plurality of dedistorted images.
[0018] Optionally, in a fourth implementation of the first aspect of the present invention, S103 includes:
[0019] S1031, based on a preset multi-view geometry method, restoring depth information of each of the object images to obtain multiple depth maps;
[0020] S1032: Fusing the multiple depth maps to generate a dense point cloud, and calculating a mesh network based on a preset TSDF algorithm.
[0021] Optionally, in a fifth implementation of the first aspect of the present invention, S105 includes:
[0022] S1051, classifying the plurality of small object images as the same object to obtain a plurality of small object image sets;
[0023] S1052: training the plurality of small object image sets one by one to obtain a plurality of small Gaussian models.
[0024] Optionally, in a sixth implementation of the first aspect of the present invention, 105 further includes:
[0025] S1053, integrating the plurality of small Gaussian models into the scene model and associating them with the mesh network;
[0026] S1054: Distribute each of the small Gaussian models to a triangle of the mesh network that is close to it, to obtain an editable Gaussian scene model.
[0027] A first implementation of the second aspect of the present invention provides a scene editing and generating system, comprising:
[0028] An acquisition module, used for acquiring a plurality of image data, and extracting and integrating a plurality of object images in each of the image data to obtain a plurality of object data sets, wherein the object images in the object data sets are of the same object;
[0029] A distortion module, used for acquiring the position and posture information of the plurality of object images, and performing dedistortion processing on the plurality of object images and the corresponding position and posture information to obtain a plurality of dedistorted images;
[0030] A restoration module, used to restore and fuse the depth information of each of the dynamic object images, and obtain a mesh network based on a preset first algorithm;
[0031] A training module, used for performing 3DGS training on the plurality of dedistorted images and the corresponding pose data to obtain a scene model, wherein the dedistorted images include small object images that do not participate in the 3DGS training;
[0032] The scene module is used to generate a plurality of small Gaussian models corresponding to the plurality of small object images, integrate the plurality of small Gaussian models into the scene model and associate them with the mesh network to obtain an editable Gaussian scene model.
[0033] A first implementation of the third aspect of the present invention provides a scene editing and generating device, the scene editing and generating device comprising: a memory and at least one processor, the memory storing instructions, the memory and the at least one processor being interconnected via a line;
[0034] The at least one processor calls the instructions in the memory to enable the scene editing and generating device to perform the scene editing and generating method as described in any one of the first aspects of the present invention.
[0035] A first implementation of the fourth aspect of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the scene editing and generation method as described in any one of the above-mentioned first aspects of the present invention is implemented.
[0036] Compared with the prior art, the present invention has the following beneficial technical effects: extracting and integrating images of the same object from multiple image data to form an object data set, obtaining the position and posture information of these object images, and performing de-distortion processing to obtain de-distorted images, restoring and fusing the depth information of dynamic object images, generating a mesh network using a preset algorithm, training the de-distorted images and their position and posture data for 3DGS, building a scene model, and processing small object images that do not participate in the training. Finally, training the small object images to obtain a small Gaussian model, integrating it into the scene model and associating it with the mesh network to form an editable Gaussian scene model, so as to improve data processing efficiency, enhance the accuracy of the scene model and realize the editability of the scene model. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] Figure 1 It is a schematic diagram of a first embodiment of a scene editing and generating method in an embodiment of the present invention;
[0038] Figure 2 Schematic diagram of a second embodiment of the scene editing and generating method in an embodiment of the present invention;
[0039] Figure 3 Schematic diagram of a third embodiment of the scene editing and generating method in an embodiment of the present invention;
[0040] Figure 4 Schematic diagram of a fourth embodiment of the scene editing and generating method in an embodiment of the present invention;
[0041] Figure 5 Schematic diagram of a fifth embodiment of a scene editing and generating method according to an embodiment of the present invention;
[0042] Figure 6 Schematic diagram of a sixth embodiment of the scene editing and generating method in an embodiment of the present invention;
[0043] Figure 7 A schematic diagram of an embodiment of a scene editing and generating system in an embodiment of the present invention;
[0044] Figure 8 A schematic diagram of an embodiment of a scene editing and generating device in an embodiment of the present invention;
[0045] Fig. 9 The following is a flowchart of the scene editing and generating method in an embodiment of the present invention. DETAILED DESCRIPTION
[0046] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.
[0047] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or component referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and the like are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, features defined as "first", "second", and the like may explicitly or implicitly include one or more features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.
[0048] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or a specific connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood by specific circumstances.
[0049] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1-Figure 6 as well as Fig. 9 In an embodiment of the present invention, a scene editing and generating method is used for detecting a road. The scene editing and generating method comprises:
[0050] S101, acquiring a plurality of image data, and extracting and integrating a plurality of object images in each of the image data to obtain a plurality of object data sets, wherein the object images in the object data sets are of the same object;
[0051] In this embodiment, object images belonging to the same object in the image data are extracted in order to subsequently help identify and classify specific targets in the image.
[0052] Furthermore, step S101 may further specifically perform:
[0053] S1011, acquiring a plurality of image data, and using a preset yolov6 algorithm to identify the plurality of image data, to obtain a plurality of object images, wherein the object images are dynamic objects;
[0054] S1012: Integrate the multiple object images belonging to the same object in the multiple image data to obtain multiple object data sets.
[0055] Specifically, the image data is a video or a photo collection, and a target detection algorithm is used to identify multiple image data to screen out the same dynamic object. In this embodiment, the target detection algorithm used is the yolov6 algorithm, which includes the COCO data set, the PASCAL VOC data set, etc., and can quickly identify multiple object images in the image data. In order to identify dynamic object images, optical flow is used for tracking processing. According to the results of optical flow tracking, image data belonging to the same object can be obtained, and they are classified and integrated into the same object data set. It should be noted that the video needs to extract key frames to reduce a certain amount of data and reduce the amount of processing to improve efficiency.
[0056] S102, obtaining position and posture information of a plurality of the object images, and performing dedistortion processing on the plurality of the object images and the corresponding position and posture information to obtain a plurality of dedistorted images;
[0057] In this embodiment, since the acquired object images have geometric distortion, in order to enable these object images to be processed and used later, it is necessary to perform dedistortion processing on these object images.
[0058] Furthermore, step S102 may further specifically execute:
[0059] S1021, using a preset COLMAP tool, processing the plurality of object data sets to obtain an SFM result, wherein the SFM result includes position and posture information of the plurality of object images;
[0060] S1022: Dedistort each of the object images and the corresponding posture information to obtain a plurality of dedistorted images.
[0061] S103, restoring and fusing the depth information of each of the dynamic object images, and obtaining a mesh network based on a preset first algorithm;
[0062] Furthermore, step S103 may further specifically execute:
[0063] S1031, based on a preset multi-view geometry method, restoring depth information of each of the object images to obtain multiple depth maps;
[0064] S1032: Fusing the multiple depth maps to generate a dense point cloud, and calculating a mesh network based on a preset TSDF algorithm.
[0065] In this embodiment, feature points or feature areas are first extracted from each object image. These features are usually parts of the image that are easy to identify and can be stably matched in different views. Then, the corresponding relationship between these features in different views is found through a matching algorithm. Using these corresponding relationships and known camera parameters (such as focal length, optical center position, relative posture between cameras, etc.), a set of equations or optimization problems about the scene depth can be constructed. By solving these sets of equations or optimization problems, the depth information of each object image can be restored, that is, the distance from each pixel to the camera plane.
[0066] The spatial consistency between the depth maps is ensured by matching the pose information and image features, so as to fuse them to obtain a dense point cloud. The TSDF algorithm divides the space into small voxels and calculates the TSDF value in each voxel. For each voxel, the TSDF value is calculated based on its distance to the nearest surface and the cutoff distance. If the voxel is in front of the surface, the TSDF value is positive; if the voxel is behind the surface, the TSDF value is negative; if the voxel is within the cutoff distance, the TSDF value is truncated, and the isosurface (i.e., surface) is extracted from the volume data through the Marching Cubes algorithm. The MarchingCubes algorithm determines which edges intersect with the isosurface by comparing the TSDF values of the voxel vertices with the user-specified threshold, and creates triangular patches to connect these edges, and finally forms a mesh network with these triangular patches.
[0067] S104, performing 3DGS training on the plurality of dedistorted images and the corresponding pose data to obtain a scene model, wherein the dedistorted images include small object images that do not participate in the 3DGS training;
[0068] In this embodiment, the small object image is the pixel cut out from the dedistorted image and does not participate in the 3DGS training. In order to achieve better training results and better generalization performance, the dense points sampled by the mesh network are used as the initial point input, and the depth map constraint and the 2D plane constraint are added so that the trained Gaussian points will not have many floating objects.
[0069] S105, training a plurality of small object images to obtain a corresponding plurality of small Gaussian models, integrating the plurality of small Gaussian models into the scene model and associating the models with the mesh network to obtain an editable Gaussian scene model.
[0070] Furthermore, step S105 may further specifically execute:
[0071] S1051, classifying the plurality of small object images as the same object to obtain a plurality of small object image sets;
[0072] S1052, training the plurality of small object image sets one by one to obtain a plurality of small Gaussian models;
[0073] In steps S1051-S1052, by associating small object images and training them separately, each independent object in the scene can obtain a refined Gaussian model representation, providing more detailed control for subsequent editing and rendering.
[0074] S1053, integrating the plurality of small Gaussian models into the scene model and associating them with the mesh network;
[0075] In this embodiment, by placing a small Gaussian model in the entire scene model and removing the Gaussian points in the corresponding range of the scene model, the effect of scene editing can be achieved. In theory, any number of small Gaussian models can be placed to achieve free editing of the scene, greatly enhancing the plasticity and personalized expression of the scene.
[0076] S1054: Distribute each of the small Gaussian models to a triangle of the mesh network that is close to it, to obtain an editable Gaussian scene model.
[0077] In this embodiment, the trained model is associated with the mesh, and each Gaussian is assigned to the closest mesh triangle. The mesh is deformed according to the specified target to achieve the required shape distribution, and the corresponding Gaussian is transformed according to the deformation of the triangle. The rotation, translation, scaling and angle change of the triangle before and after the change can be directly calculated by the vertex coordinates before and after the change of the triangle. The center point of the Gaussian needs to change according to the rotation and translation, and the scale of the Gaussian needs to be adjusted according to the scaling and angle changes. The other parameters of the Gaussian remain unchanged.
[0078] Specifically, it extracts and integrates images of the same object from multiple image data to form an object data set, obtains the pose information of these object images, and performs dedistortion processing to obtain dedistorted images. By restoring and fusing the depth information of dynamic object images, a mesh network is generated using a preset algorithm, and the dedistorted images and their pose data are trained on 3DGS to build a scene model, while processing small object images that do not participate in the training. Finally, a small Gaussian model is obtained by training small object images, which is integrated into the scene model and associated with the mesh network to form an editable Gaussian scene model to improve data processing efficiency, enhance the accuracy of the scene model, and realize the editability of the scene model.
[0079] The above describes the scene editing and generating method in the embodiment of the present invention. The following describes the scene editing and generating system in the embodiment of the present invention. Figure 7 , the scene editing and generating system comprises:
[0080] An acquisition module 201 is used to acquire a plurality of image data, and extract and integrate a plurality of object images in each of the image data to obtain a plurality of object data sets, wherein the object images in the object data sets are of the same object;
[0081] The distortion module 202 is used to obtain the position and posture information of the plurality of object images, and perform dedistortion processing on the plurality of object images and the corresponding position and posture information to obtain a plurality of dedistorted images;
[0082] A restoration module 203, used to restore and fuse the depth information of each of the dynamic object images, and obtain a mesh network based on a preset first algorithm;
[0083] A training module 204 is used to perform 3DGS training on the plurality of dedistorted images and the corresponding pose data to obtain a scene model, wherein the dedistorted images include small object images that do not participate in the 3DGS training;
[0084] The scene module 205 is used to generate a plurality of small Gaussian models corresponding to the plurality of small object images, integrate the plurality of small Gaussian models into the scene model and associate them with the mesh network to obtain an editable Gaussian scene model.
[0085] Specifically, it extracts and integrates images of the same object from multiple image data to form an object data set, obtains the pose information of these object images, and performs dedistortion processing to obtain dedistorted images. By restoring and fusing the depth information of dynamic object images, a mesh network is generated using a preset algorithm, and the dedistorted images and their pose data are trained on 3DGS to build a scene model, while processing small object images that do not participate in the training. Finally, a small Gaussian model is obtained by training small object images, which is integrated into the scene model and associated with the mesh network to form an editable Gaussian scene model to improve data processing efficiency, enhance the accuracy of the scene model, and realize the editability of the scene model.
[0086] The acquisition module 201 may also be specifically used for:
[0087] Acquire multiple image data, and use a preset yolov6 algorithm to identify the multiple image data to obtain multiple object images, wherein the object images are dynamic objects;
[0088] The object images belonging to the same object in the plurality of image data are integrated to obtain a plurality of object data sets.
[0089] The distortion module 202 may also be specifically used for:
[0090] Using a preset COLMAP tool, processing the plurality of object data sets to obtain an SFM result, wherein the SFM result includes position and posture information of the plurality of object images;
[0091] Each of the object images and the corresponding posture information is subjected to dedistortion processing to obtain a plurality of dedistorted images.
[0092] The recovery module 203 may also be specifically used for:
[0093] Based on a preset multi-view geometry method, restore the depth information of each of the object images to obtain multiple depth maps;
[0094] The multiple depth maps are fused to generate a dense point cloud, and a mesh network is calculated based on a preset TSDF algorithm.
[0095] The scene module 205 may also be specifically used for:
[0096] Classifying the plurality of small object images as the same object to obtain a plurality of small object image sets;
[0097] Training the plurality of small object image sets one by one to obtain a plurality of small Gaussian models;
[0098] Integrating a plurality of the small Gaussian models into the scene model and associating them with the mesh network;
[0099] Each of the small Gaussian models is respectively assigned to a triangle of the mesh network that is close to it, so as to obtain an editable Gaussian scene model.
[0100] above Figure 7 The scene editing and generating system in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The scene editing and generating device in the embodiment of the present invention is described in detail from the perspective of hardware processing.
[0101] Figure 83 is a schematic diagram of the structure of a scene editing and generating device provided by an embodiment of the present invention. The scene editing and generating device 300 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 310 (for example, one or more processors) and a memory 320, and one or more storage media 330 (for example, one or more mass storage devices) storing application programs 333 or data 332. Among them, the memory 320 and the storage medium 330 may be temporary storage or permanent storage. The program stored in the storage medium 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the scene editing and generating device 300. Furthermore, the processor 310 may be configured to communicate with the storage medium 330 to execute a series of instruction operations in the storage medium 330 on the scene editing and generating device 300.
[0102] The scene-based editing and generating device 300 may also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input and output interfaces 360, and / or one or more operating systems 331, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will appreciate that Figure 3 The scene editing and generation device structure shown does not constitute a limitation on the communication protocol device based on LAN screen projection, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.
[0103] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the scene editing and generation method.
[0104] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.
[0105] The above is a scene editing and generation method or multiple implementation methods provided in combination with specific content, and it is not intended that the specific implementation of the present invention is limited to these descriptions. Any method, structure, etc. similar to or identical to the present invention, or any technical deduction or replacement based on the concept of the present invention, shall be deemed to be within the scope of protection of the present invention.
Claims
1. A scene editing and generating method, characterized in that: include: S101, acquiring a plurality of image data, and extracting and integrating a plurality of object images in each of the image data to obtain a plurality of object data sets, wherein the object images in the object data sets are of the same object; S102, obtaining position and posture information of a plurality of the object images, and performing dedistortion processing on the plurality of the object images and the corresponding position and posture information to obtain a plurality of dedistorted images; S103, restoring and fusing the depth information of each of the dynamic object images, and obtaining a mesh network based on a preset first algorithm; S104, performing 3DGS training on the plurality of dedistorted images and the corresponding pose data to obtain a scene model, wherein the dedistorted images include small object images that do not participate in the 3DGS training; S105, training a plurality of small object images to obtain a corresponding plurality of small Gaussian models, integrating the plurality of small Gaussian models into the scene model and associating the models with the mesh network to obtain an editable Gaussian scene model.
2. A scene editing and generating method according to claim 1, characterized in that: The S101 includes: S1011, acquiring a plurality of image data, and using a preset yolov6 algorithm to identify the plurality of image data, to obtain a plurality of object images, wherein the object images are dynamic objects; S1012: Integrate the multiple object images belonging to the same object in the multiple image data to obtain multiple object data sets.
3. A scene editing and generating method according to claim 2, characterized in that: The S102 includes: S1021, using a preset COLMAP tool, processing the plurality of object data sets to obtain an SFM result, wherein the SFM result includes position and posture information of the plurality of object images; S1022: Dedistort each of the object images and the corresponding posture information to obtain a plurality of dedistorted images.
4. A scene editing and generating method according to claim 1, characterized in that: The S103 includes: S1031, based on a preset multi-view geometry method, restoring depth information of each of the object images to obtain multiple depth maps; S1032: Fusing the multiple depth maps to generate a dense point cloud, and calculating a mesh network based on a preset TSDF algorithm.
5. A scene editing and generating method according to claim 1, characterized in that: The S105 includes: S1051, classifying the plurality of small object images as the same object to obtain a plurality of small object image sets; S1052: training the plurality of small object image sets one by one to obtain a plurality of small Gaussian models.
6. A scene editing and generating method according to claim 5, characterized in that: The 105 also includes: S1053, integrating the plurality of small Gaussian models into the scene model and associating them with the mesh network; S1054: Distribute each of the small Gaussian models to a triangle of the mesh network that is close to it, to obtain an editable Gaussian scene model.
7. A scene editing and generation system, characterized in that: include: An acquisition module, used for acquiring a plurality of image data, and extracting and integrating a plurality of object images in each of the image data to obtain a plurality of object data sets, wherein the object images in the object data sets are of the same object; A distortion module, used for acquiring the position and posture information of the plurality of object images, and performing dedistortion processing on the plurality of object images and the corresponding position and posture information to obtain a plurality of dedistorted images; A restoration module, used to restore and fuse the depth information of each of the dynamic object images, and obtain a mesh network based on a preset first algorithm; A training module, used for performing 3DGS training on the plurality of dedistorted images and the corresponding pose data to obtain a scene model, wherein the dedistorted images include small object images that do not participate in the 3DGS training; The scene module is used to generate a plurality of small Gaussian models corresponding to the plurality of small object images, integrate the plurality of small Gaussian models into the scene model and associate them with the mesh network to obtain an editable Gaussian scene model.
8. A scene editing and generating device, characterized in that: The scene editing and generating device comprises: a memory and at least one processor, the memory stores instructions, and the memory and the at least one processor are interconnected via a line; The at least one processor calls the instructions in the memory to enable the scene editing and generating device to perform the scene editing and generating method according to any one of claims 1 to 6.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the scene editing and generating method according to any one of claims 1 to 6 is implemented.