A VR-based regional landscape immersive evaluation prediction method and device

Abstract

The application discloses a VR-based regional landscape immersive evaluation prediction method and device, which comprises the following steps: acquiring high-resolution multi-view image data of a target region, combining a motion recovery structure with a multi-view stereo vision algorithm to construct an original three-dimensional real scene model; then, through grid simplification and space index processing, multi-detail level streaming three-dimensional data is generated, relying on a virtual reality platform to fuse terrain data, configure rendering parameters, build an interactive virtual regional environment and adapt to VR equipment; semantic score data and auxiliary evaluation data are acquired; the semantic score data and the auxiliary evaluation data are input into a pre-constructed overall landscape satisfaction prediction model to obtain a satisfaction prediction score; and the application relies on three-dimensional real scenes and VR interaction to truly restore regional landscape features; and by combining auxiliary evaluation and quantitative analysis, qualitative landscape evaluation is converted into quantitative landscape evaluation.

Description

Technical Field

[0001] This invention relates to a VR-based immersive evaluation and prediction method and device for regional landscapes, belonging to the field of virtual reality technology. Background Technology

[0002] As a vital spatial carrier of regional culture, historical memory, and local sentiment, the scientific and systematic evaluation of regional landscape quality has become a crucial foundation for rural planning, tourism development, and the improvement of the living environment. Accurate landscape evaluation not only concerns the recognition of aesthetic value but also directly impacts cultural inheritance, ecological protection, and the establishment of social identity.

[0003] Currently, commonly used methods for regional landscape evaluation include beauty score evaluation, comparative evaluation, and semantic analysis, primarily using two-dimensional static images and videos as evaluation media. However, traditional media have inherent limitations in expressing spatial continuity, field of vision, and depth perception, easily leading to perceptual biases, and also resulting in information loss in reproducing spatial depth, dynamic light and shadow, and environmental details. Although scholars have attempted to introduce virtual reality (VR) technology in recent years, existing research largely relies on 360° panoramic images from fixed locations, and the evaluation dimensions remain at the level of static visual perception, making it difficult to reproduce the complex topographic features and rich spatial layers of regional landscapes. Summary of the Invention

[0004] The technical problem to be solved by this invention is that the existing technology for regional landscape evaluation relies on two-dimensional static images, which makes it difficult to reproduce the true spatial perception.

[0005] To solve the above-mentioned technical problems, the present invention is implemented by a VR-based immersive evaluation and prediction method, device and system for regional landscapes.

[0006] In a first aspect, the present invention provides a VR-based immersive evaluation and prediction method for regional landscapes, comprising:

[0007] Acquire high-resolution multi-view image data of the target area;

[0008] Based on the high-resolution multi-view image data of the region, a high-precision original 3D real scene model is generated by using multi-view image motion recovery structure and multi-view stereo vision algorithm.

[0009] The original 3D reality model is simplified by meshing and spatial indexing to generate streaming 3D model data that supports dynamic loading of multiple levels of detail.

[0010] In the virtual reality development platform, the streaming 3D model data is spatially matched and fused with global terrain data, rendering parameters are configured, an interactive virtual area environment is constructed, and it is compiled and published to immersive VR devices.

[0011] Semantic scoring data and auxiliary evaluation data are obtained. The semantic scoring data is data collected after the VR device performs semantic scoring on multiple landscape feature factors of the virtual area environment. The auxiliary evaluation data is data collected after inputting the regional spherical panoramic image and multi-view slices from the high-resolution multi-view image data of the area into a large language model for auxiliary evaluation.

[0012] The semantic scoring data and auxiliary evaluation data are input into a pre-built overall landscape satisfaction prediction model to obtain a satisfaction prediction score.

[0013] The construction of the overall landscape satisfaction prediction model includes: acquiring historical semantic score data and historical auxiliary evaluation data of multiple areas in the target type region; standardizing the historical semantic score data and historical auxiliary evaluation data; performing correlation analysis and multiple regression analysis on the standardized data to screen out key landscape influencing factors; and constructing an overall landscape satisfaction prediction model based on the key landscape influencing factors.

[0014] This invention uses a multi-view image motion recovery structure and a multi-view stereo vision algorithm to generate a high-precision original 3D real-world model. Then, it uses mesh simplification and spatial indexing to ensure the model's fidelity and detail accuracy. VR devices are used to perform semantic scoring on the virtual environment, improving the intuitiveness and realism of landscape evaluation.

[0015] The target type of region includes plains, hills, or other types of regions, where other types of regions are regions other than plains and hills.

[0016] The steps for acquiring high-resolution multi-view image data of the target area include: using a drone equipped with an oblique photography camera to conduct multiple aerial photographs of the study area from multiple angles, adopting a crossfire network-style five-way flight route, setting the camera gimbal tilt angle, and planning its directional overlap and lateral overlap to acquire high-resolution oblique image data of the area.

[0017] It can collect multi-angle tilted images of the area from all directions without blind spots, effectively improving the integrity and coverage of image data.

[0018] The steps of employing multi-view image motion recovery structure and multi-view stereo vision algorithm include:

[0019] The high-resolution multi-view image data of the region is used to extract feature points for aerial triangulation using the multi-view image motion recovery structure algorithm. The formula for aerial triangulation is as follows:

[0020]

[0021]

[0022] in, The horizontal axis coordinates of the image points. This refers to the horizontal orientation element within the camera. , , Let be a rotation matrix composed of exterior orientation elements. , For camera principal distance, , , These are the object coordinates corresponding to the image point. , , For image exterior orientation elements, The vertical coordinates of the image points are: This refers to the vertical orientation element within the camera;

[0023] By combining multi-view stereo vision algorithms to generate dense point clouds and irregular triangular meshes, a high-precision original 3D reality model in OSGB format is generated.

[0024] By employing multi-view image motion reconstruction structure and multi-view stereo vision algorithm, the geometric shape and spatial relationship of regional buildings, vegetation and terrain are accurately restored.

[0025] The steps of simplifying the mesh and constructing the spatial index of the original 3D reality model include:

[0026] The high-precision original 3D reality model in OSGB format is preprocessed using a quadratic error metric grid simplification algorithm.

[0027] The simplified high-precision original 3D reality model is spatially sliced ​​and graded using KD-trees to construct a spatial index and generate 3D Tiles format 3D model data that supports dynamic loading of multiple levels of detail.

[0028] The QEM mesh simplification algorithm is combined with KD tree spatial slicing and spatial indexing to improve the smoothness of virtual scene operation and meet the stable operation requirements of VR devices.

[0029] The steps for constructing an interactive, roamable virtual area environment include:

[0030] In the Unity development platform, configure the 3D terrain plugin and load global terrain data;

[0031] The 3D Tiles format 3D model data is stored on a local server built on XAMPP and loaded into the Unity scene via file path.

[0032] The 3D terrain plugin automatically matches and fits the regional model with the global terrain based on the latitude, longitude and elevation information embedded in the 3D Tiles format 3D model.

[0033] Adjust the maximum screen space error parameter of the virtual area scene.

[0034] Construct virtual environments that are realistic in scale, spatially complete, and freely roamable to enhance the spatial realism and interactive smoothness of VR roaming.

[0035] The various landscape characteristic factors include: color harmony, vegetation richness, vegetation distribution pattern, building layout harmony, environmental comfort, environmental order, field of vision, overall harmony, distinctiveness of rural features, and sense of spatial hierarchy.

[0036] The steps for assisting evaluation using the large language model include:

[0037] Construct a structured prompt that includes four main modules: character setting, task background, indicator definition, and scoring rules;

[0038] The spherical panoramic image and multi-view slices of the region are input into a large language model to obtain a systematic evaluation reference based on the landscape feature factors.

[0039] The steps for screening key landscape influencing factors through correlation analysis and multiple regression analysis include:

[0040] Calculate the Pearson correlation coefficient between overall landscape satisfaction and each landscape characteristic factor, conduct correlation analysis, and preliminarily screen out factors that are significantly related to overall satisfaction.

[0041] Using overall satisfaction as the dependent variable and the selected landscape characteristic factors as independent variables, a stepwise multiple linear regression method was adopted. Based on the results of analysis of variance and significance test, variables with multicollinearity were eliminated, and finally a linear predictive regression equation for overall landscape satisfaction was established.

[0042] Pearson correlation analysis was used to initially screen effective landscape factors, and stepwise multiple linear regression was used to eliminate multicollinear variables, thereby improving the scientific rigor and reliability of the model.

[0043] Secondly, the present invention provides a VR-based immersive evaluation and prediction device for regional landscapes, comprising:

[0044] The image data acquisition module is used to acquire high-resolution multi-view image data of the target area;

[0045] The 3D reality model generation module is used to: generate a high-precision original 3D reality model based on the high-resolution multi-view image data of the region, using multi-view image motion recovery structure and multi-view stereo vision algorithm;

[0046] The 3D reality model processing module is used to: simplify the mesh and construct the spatial index of the original 3D reality model to generate streaming 3D model data that supports dynamic loading of multiple levels of detail;

[0047] The virtual area environment construction module is used to: spatially match and fuse the streaming 3D model data with global terrain data in the virtual reality development platform, configure rendering parameters, construct an interactive virtual area environment, and compile and publish it to immersive VR devices.

[0048] The evaluation data acquisition module is used to: acquire semantic scoring data and auxiliary evaluation data. The semantic scoring data is data collected in advance after using the VR device to perform semantic scoring on multiple landscape feature factors of the virtual area environment. The auxiliary evaluation data is data collected after inputting the regional spherical panoramic image and multi-view slices from the high-resolution multi-view image data of the region into a large language model for auxiliary evaluation.

[0049] The regional environmental assessment module is used to: input the semantic scoring data and auxiliary evaluation data into a pre-built overall landscape satisfaction prediction model to obtain a satisfaction prediction score;

[0050] The overall landscape satisfaction prediction model is constructed by: acquiring historical semantic score data and historical auxiliary evaluation data of multiple areas in the target type region; standardizing the historical semantic score data and historical auxiliary evaluation data; performing correlation analysis and multiple regression analysis on the standardized data to screen out key landscape influencing factors; and constructing an overall landscape satisfaction prediction model based on the key landscape influencing factors.

[0051] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0052] This invention breaks through the limitations of traditional on-site surveys and static image evaluations of regional landscapes by constructing realistic 3D models and VR immersive interaction, truly restoring the overall appearance of the region and realizing an immersive and visual experience for landscape evaluation; by constructing a satisfaction prediction model through data statistical analysis, it realizes the transformation of regional landscape evaluation from qualitative description to quantitative analysis, accurately explores the laws of landscape impact, and provides scientific and objective technical support for regional planning and design. Attached Figure Description

[0053] Figure 1 This is a flowchart illustrating the VR-based immersive evaluation and prediction method for regional landscapes as shown in Embodiment 1 of the present invention.

[0054] Figure 2 This is the original three-dimensional real-scene model of the sample area shown in Embodiment 2 of the present invention;

[0055] Figure 3 This is a village scene in the PICO VR device shown in Embodiment 2 of the present invention;

[0056] Figure 4 This is a diagram showing the overall satisfaction with the village environment, as illustrated in Embodiment 2 of the present invention. Detailed Implementation

[0057] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0058] The term "and / or" simply describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. Additionally, the character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0059] Example 1

[0060] like Figure 1 This embodiment introduces a VR-based immersive evaluation method for regional landscapes, including:

[0061] Acquire high-resolution multi-view image data of the target area;

[0062] Based on regional high-resolution multi-view image data, a high-precision original 3D real scene model is generated by using multi-view image motion recovery structure and multi-view stereo vision algorithm.

[0063] The original 3D reality model is simplified by meshing and spatial indexing to generate streaming 3D model data that supports dynamic loading of multiple levels of detail.

[0064] In the virtual reality development platform, streaming 3D model data is spatially matched and fused with global terrain data, rendering parameters are configured, an interactive virtual area environment is constructed, and it is compiled and published to immersive VR devices.

[0065] Semantic scoring data and auxiliary evaluation data were acquired. The semantic scoring data was collected after the virtual area environment was semantically scored using VR devices for multiple landscape feature factors. The auxiliary evaluation data was collected after the regional spherical panoramic image and multi-view slices from the regional high-resolution multi-view image data were input into a large language model for auxiliary evaluation.

[0066] The semantic score data and auxiliary evaluation data are input into a pre-built overall landscape satisfaction prediction model to obtain the satisfaction prediction score;

[0067] The construction of the overall landscape satisfaction prediction model includes: acquiring historical semantic score data and historical auxiliary evaluation data of multiple regions in the target type area; standardizing the historical semantic score data and historical auxiliary evaluation data; conducting correlation analysis and multiple regression analysis on the standardized data to screen out key landscape influencing factors; and constructing an overall landscape satisfaction prediction model based on the key landscape influencing factors.

[0068] If the target area is a region other than plains and hilly areas, after obtaining semantic score data and auxiliary evaluation data, a new overall landscape satisfaction prediction model is constructed to obtain the satisfaction prediction score.

[0069] The construction steps for the new overall landscape satisfaction prediction model are the same as those for the overall landscape satisfaction prediction model.

[0070] The steps for acquiring high-resolution multi-view image data of the target area include: using a drone equipped with an oblique photography camera to conduct multiple sorties and multi-angle aerial photography of the study area, adopting a crossfire network five-way flight route, setting the camera gimbal tilt angle, and planning its directional overlap and lateral overlap to acquire high-resolution oblique image data of the area.

[0071] Specifically, the five-directional flight route is: one orthophoto and four multi-angle oblique images are collected together, with the multi-angle images usually being the four side views of the east, south, west and north.

[0072] The steps of using multi-view image motion reconstruction structure and multi-view stereo vision algorithm include:

[0073] Aerial triangulation is performed on regional high-resolution multi-view image data using a multi-view image motion reconstruction structure algorithm to extract feature points. The formula for aerial triangulation is as follows:

[0074]

[0075]

[0076] in, The horizontal axis coordinates of the image points. This refers to the horizontal orientation element within the camera. , , Let be a rotation matrix composed of exterior orientation elements. , For camera principal distance, , , These are the object coordinates corresponding to the image point. , , For image exterior orientation elements, The vertical coordinates of the image points are: This refers to the vertical orientation element within the camera;

[0077] By combining multi-view stereo vision algorithms to generate dense point clouds and irregular triangular meshes, a high-precision original 3D reality model in OSGB format is generated.

[0078] Specifically, the multi-view stereo vision algorithm includes: calculating the pixel-level depth value of the image using multi-view geometric constraints, performing fusion processing on the generated depth map to obtain a dense point cloud reflecting the high-density surface information of the regional landscape, and performing noise reduction and filtering processing.

[0079] The steps for mesh simplification and spatial indexing of the original 3D reality model include:

[0080] A quadratic error metric grid simplification algorithm is used to preprocess the high-precision original 3D reality model in OSGB format;

[0081] The simplified high-precision original 3D reality model is spatially sliced ​​and graded using KD-trees to construct a spatial index and generate 3D Tiles format 3D model data that supports dynamic loading of multiple levels of detail.

[0082] Specifically, the quadratic error metric grid simplification algorithm is based on the Heckbert-Garland algorithm logic, which calculates the quadratic error cost function generated by the folded edges.

[0083] The steps to construct an interactive, roamable virtual area environment include:

[0084] In the Unity development platform, configure the 3D terrain plugin and load global terrain data;

[0085] The 3D Tiles format 3D model data is stored on a local server built on XAMPP and loaded into the Unity scene via file path.

[0086] The 3D terrain plugin automatically matches and fits the regional model with the global terrain based on the latitude, longitude and elevation information embedded in the 3D Tiles format 3D model.

[0087] Adjust the maximum screen space error parameter for the virtual area scene.

[0088] Specifically, the maximum screen space error parameter is set as follows: the loading level of 3D Tiles is dynamically controlled based on the viewpoint distance. When the screen space error exceeds the set threshold, the system automatically calls high-level (LOD) tiles to ensure the sharpness of the foreground, and vice versa, low-level tiles are called to maintain the running frame rate.

[0089] Several landscape characteristic factors include: color harmony, vegetation richness, vegetation distribution pattern, building layout harmony, environmental comfort, environmental order, field of vision, overall harmony, distinctiveness of rural features, and sense of spatial hierarchy.

[0090] The steps for large language models to perform auxiliary evaluation include:

[0091] Construct a structured prompt that includes four main modules: character setting, task background, indicator definition, and scoring rules;

[0092] By inputting the spherical panoramic image and multi-view slices of the region into a large language model, a systematic evaluation reference based on landscape feature factors is obtained.

[0093] The steps for screening key landscape influencing factors through correlation analysis and multiple regression analysis include:

[0094] Calculate the Pearson correlation coefficient between overall landscape satisfaction and each landscape characteristic factor, conduct correlation analysis, and preliminarily screen out factors that are significantly related to overall satisfaction.

[0095] Using overall satisfaction as the dependent variable and the selected landscape characteristic factors as independent variables, a stepwise multiple linear regression method was adopted. Based on the results of analysis of variance and significance test, variables with multicollinearity were eliminated, and finally a linear predictive regression equation for overall landscape satisfaction was established.

[0096] Specifically, the correlation analysis includes: calculating the correlation coefficient between overall satisfaction and each landscape factor, initially identifying key influencing factors, excluding elements that have no significant impact on the regional landscape, and screening out landscape factors that have a significant linear correlation with overall satisfaction.

[0097] Specifically, the linear predictive regression equation for overall landscape satisfaction is established by using a stepwise multiple linear regression method, with overall satisfaction as the dependent variable and color harmony, vegetation richness, vegetation distribution pattern, building layout harmony, environmental comfort, environmental order, field of vision, overall harmony, distinctiveness of rural features, and spatial hierarchy as independent variables. Based on the results of variance analysis and significance test in the software, multicollinear variables are eliminated, and finally, the regression equation is established.

[0098] Example 2

[0099] For example Figure 2 The sample area shown adopts a VR-based immersive regional landscape evaluation and prediction method, and the steps are as follows:

[0100] Step 1: Use the DJI Phantom 4 RTK drone to conduct multiple sorties and multi-angle aerial photography of 16 typical mountain villages in Jinzhai County, Anhui Province. Use a five-way flight path, set the camera gimbal tilt angle to -45° to -60°, plan 80% forward overlap and 75% lateral overlap, and obtain high-resolution oblique image data of the villages.

[0101] Step 2: Use Bentley Context Capture real-scene modeling software to process the acquired aerial images, use the multi-view image motion recovery structure algorithm to extract feature points for aerial triangulation calculation, and combine it with multi-view stereo vision algorithm to generate dense point cloud and irregular triangular mesh, generate original 3D real-scene model in OSGB format, and ensure that the overall accuracy of the model is better than 4 cm.

[0102] Step 3: Preprocess the generated OSGB format original 3D model using Super Map iDesktop software, introducing a quadratic error metric mesh simplification algorithm to reduce the number of polygons while preserving the building outline. Then, use a KD tree (K-Dimensional Tree) for spatial slicing and hierarchical classification to generate a 3DTiles format model that supports dynamic loading of multiple levels of detail.

[0103] Step 4: Integrate Cesium for Unity, Cesium World Terrain, and Cesium3Dtileset plugins into the Unity 3D development platform. Then load global terrain data and the standardized 3D Tiles data obtained in Step 3. Use the Cesium Georeference component to input the actual latitude and longitude coordinates and elevation information of the village to achieve accurate spatial matching between the real-world 3D model and the terrain data, and construct a freely roamable virtual village scene that includes terrain undulations.

[0104] Step 5: Package and publish the completed Unity project as an Android APK application by installing the PICO SDK, setting up the head-mounted display and controller interaction components, and then install it on the PICO VR all-in-one device.

[0105] Step six, as follows Figure 3 As shown, immersive roaming evaluation was conducted based on PICO VR devices, and large language models were used for auxiliary evaluation. Five-level semantic scoring data and auxiliary evaluation data for 10 landscape feature factors such as color harmony, vegetation richness, and field of view were collected.

[0106] Step 7: First, based on the scoring data obtained in Step 6, calculate the mean and standard deviation of overall satisfaction and 10 landscape characteristic factors for the 16 villages. Then, calculate the correlation coefficient between overall satisfaction and each landscape factor to preliminarily identify key influencing factors, exclude elements that have no significant impact on the rural landscape, and screen out landscape factors that have a significant linear correlation with overall satisfaction. Finally, enter the data into SPSS 26.0 statistical software. Subsequently, use correlation analysis to screen significant factors, and use stepwise multiple linear regression analysis to obtain the key landscape influencing factors of vegetation distribution pattern and field of view openness, and construct a landscape satisfaction prediction model.

[0107] The scores for vegetation distribution pattern and field of view openness from semantic scoring data and auxiliary evaluation data are input into the landscape satisfaction prediction model to obtain the overall predicted value of rural landscape satisfaction, such as... Figure 4 As shown, the regression equation formula is:

[0108]

[0109] in, This represents the predicted overall satisfaction value for rural landscapes based on VR. Score the vegetation distribution pattern. Score the field of vision.

[0110] Example 3

[0111] Based on the same inventive concept as other embodiments, this embodiment introduces a VR-based immersive evaluation and prediction device for regional landscapes, comprising:

[0112] The image data acquisition module is used to acquire high-resolution multi-view image data of the target area;

[0113] The 3D reality model generation module is used to generate high-precision original 3D reality models based on regional high-resolution multi-view image data, using multi-view image motion recovery structure and multi-view stereo vision algorithm.

[0114] The 3D reality model processing module is used to: simplify the mesh and construct the spatial index of the original 3D reality model, and generate streaming 3D model data that supports dynamic loading of multiple levels of detail;

[0115] The virtual area environment construction module is used to: spatially match and fuse streaming 3D model data with global terrain data in a virtual reality development platform, configure rendering parameters, construct an interactive and roamable virtual area environment, and compile and publish it to immersive VR devices.

[0116] The evaluation data acquisition module is used to: acquire semantic scoring data and auxiliary evaluation data. The semantic scoring data is the data collected in advance after using VR devices to perform semantic scoring of multiple landscape feature factors on the virtual area environment. The auxiliary evaluation data is the data collected after inputting the regional spherical panoramic image and multi-view slices from the regional high-resolution multi-view image data into a large language model for auxiliary evaluation.

[0117] The regional environmental assessment module is used to input semantic scoring data and auxiliary assessment data into a pre-built overall landscape satisfaction prediction model to obtain a satisfaction prediction score.

[0118] The overall landscape satisfaction prediction model is constructed by: acquiring historical semantic score data and historical auxiliary evaluation data of multiple areas in the target type region; standardizing the historical semantic score data and historical auxiliary evaluation data; performing correlation analysis and multiple regression analysis on the standardized data to screen out key landscape influencing factors; and constructing an overall landscape satisfaction prediction model based on the key landscape influencing factors.

[0119] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0120] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0121] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0122] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0123] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A VR-based immersive evaluation and prediction method for regional landscapes, characterized in that, include: Acquire high-resolution multi-view image data of the target area; Based on the high-resolution multi-view image data of the region, a high-precision original 3D real scene model is generated by using multi-view image motion recovery structure and multi-view stereo vision algorithm. The original 3D reality model is simplified by meshing and spatial indexing to generate streaming 3D model data that supports dynamic loading of multiple levels of detail. In the virtual reality development platform, the streaming 3D model data is spatially matched and fused with global terrain data, rendering parameters are configured, an interactive virtual area environment is constructed, and it is compiled and published to immersive VR devices. Semantic scoring data and auxiliary evaluation data are obtained. The semantic scoring data is data collected after the VR device performs semantic scoring on multiple landscape feature factors of the virtual area environment. The auxiliary evaluation data is data collected after inputting the regional spherical panoramic image and multi-view slices from the high-resolution multi-view image data of the area into a large language model for auxiliary evaluation. The semantic scoring data and auxiliary evaluation data are input into a pre-built overall landscape satisfaction prediction model to obtain a satisfaction prediction score. The construction of the overall landscape satisfaction prediction model includes: acquiring historical semantic score data and historical auxiliary evaluation data of multiple areas in the target type region; standardizing the historical semantic score data and historical auxiliary evaluation data; performing correlation analysis and multiple regression analysis on the standardized data to screen out key landscape influencing factors; and constructing an overall landscape satisfaction prediction model based on the key landscape influencing factors.

2. The VR-based immersive evaluation and prediction method for regional landscapes according to claim 1, characterized in that, The target type of region includes plains, hills, or other types of regions, where other types of regions are regions other than plains and hills.

3. The VR-based immersive evaluation and prediction method for regional landscapes according to claim 1, characterized in that, The steps for acquiring high-resolution multi-view image data of the target area include: using a drone equipped with an oblique photography camera to conduct multiple aerial photographs of the study area from multiple angles, adopting a crossfire network-style five-way flight route, setting the camera gimbal tilt angle, and planning its directional overlap and lateral overlap to acquire high-resolution oblique image data of the area.

4. The VR-based immersive evaluation and prediction method for regional landscapes according to claim 1, characterized in that, The steps of employing multi-view image motion recovery structure and multi-view stereo vision algorithm include: The high-resolution multi-view image data of the region is used to extract feature points for aerial triangulation using the multi-view image motion recovery structure algorithm. The formula for aerial triangulation is as follows: in, The horizontal axis coordinates of the image points. This refers to the horizontal orientation element within the camera. , , Let be a rotation matrix composed of exterior orientation elements. , For camera principal distance, , , These are the object coordinates corresponding to the image point. , , For image exterior orientation elements, The vertical coordinates of the image points are: This refers to the vertical orientation element within the camera; By combining multi-view stereo vision algorithms to generate dense point clouds and irregular triangular meshes, a high-precision original 3D reality model in OSGB format is generated.

5. The VR-based immersive evaluation and prediction method for regional landscapes according to claim 1, characterized in that, The steps of simplifying the mesh and constructing the spatial index of the original 3D reality model include: The high-precision original 3D reality model in OSGB format is preprocessed using a quadratic error metric grid simplification algorithm. The simplified high-precision original 3D reality model is spatially sliced ​​and graded using KD-trees to construct a spatial index and generate 3D Tiles format 3D model data that supports dynamic loading of multiple levels of detail.

6. The VR-based immersive evaluation and prediction method for regional landscapes according to claim 1, characterized in that, The steps for constructing an interactive, roamable virtual area environment include: In the Unity development platform, configure the 3D terrain plugin and load global terrain data; The 3D Tiles format 3D model data is stored on a local server built on XAMPP and loaded into the Unity scene via file path. The 3D terrain plugin automatically matches and fits the regional model with the global terrain based on the latitude, longitude and elevation information embedded in the 3D Tiles format 3D model. Adjust the maximum screen space error parameter of the virtual area scene.

7. The VR-based immersive evaluation and prediction method for regional landscapes according to claim 1, characterized in that, The various landscape characteristic factors include: color harmony, vegetation richness, vegetation distribution pattern, building layout harmony, environmental comfort, environmental order, field of vision, overall harmony, distinctiveness of rural features, and sense of spatial hierarchy.

8. The VR-based immersive evaluation and prediction method for regional landscapes according to claim 1, characterized in that, The steps for assisting evaluation using the large language model include: Construct a structured prompt that includes four main modules: character setting, task background, indicator definition, and scoring rules; The spherical panoramic image and multi-view slices of the region are input into a large language model to obtain a systematic evaluation reference based on the landscape feature factors.

9. The VR-based immersive evaluation and prediction method for regional landscapes according to claim 1, characterized in that, The step of screening key landscape influencing factors by performing correlation analysis and multiple regression analysis on the standardized data includes: Calculate the Pearson correlation coefficient between overall landscape satisfaction and each landscape characteristic factor, conduct correlation analysis, and preliminarily screen out factors that are significantly related to overall satisfaction. Using overall satisfaction as the dependent variable and the selected landscape characteristic factors as independent variables, a stepwise multiple linear regression method was adopted. Based on the results of analysis of variance and significance test, variables with multicollinearity were eliminated, and finally a linear predictive regression equation for overall landscape satisfaction was established.

10. A VR-based immersive evaluation and prediction device for regional landscapes, characterized in that, include: The image data acquisition module is used to acquire high-resolution multi-view image data of the target area; The 3D reality model generation module is used to: generate a high-precision original 3D reality model based on the high-resolution multi-view image data of the region, using multi-view image motion recovery structure and multi-view stereo vision algorithm; The 3D reality model processing module is used to: simplify the mesh and construct the spatial index of the original 3D reality model to generate streaming 3D model data that supports dynamic loading of multiple levels of detail; The virtual area environment construction module is used to: spatially match and fuse the streaming 3D model data with global terrain data in the virtual reality development platform, configure rendering parameters, construct an interactive virtual area environment, and compile and publish it to immersive VR devices. The evaluation data acquisition module is used to: acquire semantic scoring data and auxiliary evaluation data. The semantic scoring data is data collected in advance after using the VR device to perform semantic scoring on multiple landscape feature factors of the virtual area environment. The auxiliary evaluation data is data collected after inputting the regional spherical panoramic image and multi-view slices from the high-resolution multi-view image data of the region into a large language model for auxiliary evaluation. The regional environmental assessment module is used to: input the semantic scoring data and auxiliary evaluation data into a pre-built overall landscape satisfaction prediction model to obtain a satisfaction prediction score; The overall landscape satisfaction prediction model is constructed by: acquiring historical semantic score data and historical auxiliary evaluation data of multiple areas in the target type region; standardizing the historical semantic score data and historical auxiliary evaluation data; performing correlation analysis and multiple regression analysis on the standardized data to screen out key landscape influencing factors; and constructing an overall landscape satisfaction prediction model based on the key landscape influencing factors.

Images