Large-scale point cloud rendering system and method based on multi-scale detail model
Through a multi-scale detail model system, KD-Tree and octree are used for efficient combination and segmentation of point clouds, and dynamically adjusting rendering details with coefficient hierarchical tables, solving the problem of details loss in traditional rendering methods, achieving high efficiency and quality Good large-scale point cloud rendering.
Patent Information
- Application Number
- CN202411017413.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-29
- Publication Date
- 2024-11-29
AI Technical Summary
Traditional large-scale point cloud rendering methods may lose important details when dealing with point clouds with rich details, resulting in poor rendering quality and unable to meet the needs of real-time rendering.
Using a system based on multi-scale detail model, efficient combination and segmentation of point clouds is performed through KD-Tree and octree data structures, density points are calculated as representative points, and a coefficient hierarchical table is established based on the three-dimensional scene scaling coefficients, and the point cloud is switched dynamically. Detail level is rendered, combined with KD-Tree and octree for point cloud downsampling and division of hierarchies, preserving structural features and details.
It significantly improves rendering efficiency and quality, meets the needs of real-time rendering, avoids the loss of detailed information, and provides high-quality point cloud rendering effects and a good visual experience.
Smart Images

Figure CN119048666A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of point cloud rendering, and in particular to a large-scale point cloud rendering system and method based on a multi-scale detail model. Background Art
[0002] With the rapid development of 3D modeling, virtual reality, autonomous driving, high-precision maps and other fields, point cloud technology has been widely used. Point cloud data can accurately represent the surface shape and spatial position information of objects, providing important basic data for these fields.
[0003] However, as the scale of point cloud data continues to increase, traditional point cloud rendering methods face many challenges. In the rendering process of large-scale point clouds, massive amounts of data need to be processed, which places extremely high demands on the performance of the rendering system. The direct rendering method draws all points in the point cloud into the three-dimensional scene. This method does not change the number of point clouds at different scales, resulting in low rendering efficiency and cannot meet the needs of real-time rendering.
[0004] Hierarchical rendering methods improve rendering efficiency by establishing a hierarchical structure of point clouds, but when processing point clouds with rich details, important detail information may be lost, thus affecting the rendering quality. The shortcomings of these existing technologies limit the effect and experience of large-scale point clouds in practical applications.
[0005] Therefore, in order to meet the needs of real-time rendering and display of large-scale point clouds at different scales while ensuring rendering quality and efficiency, an innovative large-scale point cloud rendering system and method based on a multi-scale detail model is urgently needed. Summary of the invention
[0006] In order to solve the problem that traditional large-scale point cloud rendering may lose important detail information when processing point clouds with rich details, thereby affecting the rendering quality, the purpose of the present invention is to provide a large-scale point cloud rendering system and method based on a multi-scale detail model.
[0007] In order to achieve the above object, the present invention adopts the following technical solution: a large-scale point cloud rendering system based on a multi-scale detail model, comprising:
[0008] The point cloud reading module is used to read the point cloud file and determine the number of scales to be divided according to the number of points in the point cloud. The calculation formula of the number of scales num of the divided point cloud is: num=total / min, where total is the number of points in the point cloud and min is the minimum number of points to be displayed;
[0009] Data combination and segmentation module, including:
[0010] KD-Tree data structure unit, used to efficiently combine point cloud data to perform nearest neighbor search of points;
[0011] The octree segmentation unit is used to segment the point cloud and iteratively divide the point cloud into num point clouds. The specific process is as follows:
[0012] The point cloud is stored in the KD-Tree data structure for subsequent representative point calculation;
[0013] Select a scale and iteratively divide the point cloud using an octree structure until the number of cells divided is the same as the number of points corresponding to the scale;
[0014] Use the DBSCAN algorithm to calculate the density points of the point cloud in the cell. During the calculation process, the nearest neighbor search is performed on each point to find the point with the highest density as the representative of the cell;
[0015] A coefficient hierarchy table establishment module is used to establish a system hierarchy table according to the three-dimensional scene scaling coefficient, where the scaling coefficient is divided according to the scale number in the range of 0-500. When the scaling coefficient exceeds 500, the lowest level detail point cloud is used for rendering and display;
[0016] The real-time monitoring and rendering module is used to monitor the orthogonal projection coefficients of the three-dimensional scene camera in real time when the user performs interactive operations such as zooming and panning, and to instantly query the hierarchy table based on this, dynamically switch to the most suitable point cloud detail level for rendering, and after the interactive operation is completed, the system immediately returns to the initial full-detail point cloud display state.
[0017] Preferably, in the data combination and segmentation module, the point cloud is downsampled and divided into levels by combining the KD-Tree data structure and the octree, which can better preserve the structural features and details of the point cloud.
[0018] Preferably, in the coefficient hierarchy table establishment module, dynamic updating of multi-scale details is achieved by inputting the number of point clouds and camera coefficients to establish a coefficient hierarchy table, so as to meet the point cloud detail display requirements at different observation distances.
[0019] A large-scale point cloud rendering method based on a multi-scale detail model comprises the following steps:
[0020] Step 1: Point cloud reading and scale determination, read the point cloud file, determine the number of scales divided according to the number of points in the point cloud, the point cloud scale number num calculation formula is: num = total / min, where total is the number of points in the point cloud, and min is the minimum number of points to be displayed;
[0021] Step 2: Data combination and segmentation, including:
[0022] Use KD-Tree data structure to efficiently combine point cloud data for nearest neighbor search of points;
[0023] Use the octree to segment the point cloud and iteratively divide the point cloud into num point clouds, including:
[0024] The point cloud is stored in the KD-Tree data structure for subsequent representative point calculation;
[0025] Select a scale and iteratively divide the point cloud using an octree structure until the number of cells divided is the same as the number of points corresponding to the scale;
[0026] Use the DBSCAN algorithm to calculate the density points of the point cloud in the cell, and find the point with the highest density as the representative of the cell by performing a nearest neighbor search on each point;
[0027] Step 3: Establish a coefficient hierarchy table. Establish a system hierarchy table based on the 3D scene scaling coefficient. The scaling coefficient is divided according to the scale quantity in the range of 0-500. When the scaling coefficient exceeds 500, the lowest level detail point cloud is used for rendering and display.
[0028] Step 4: Real-time monitoring and rendering. When the user performs interactive operations such as zooming and panning, the orthogonal projection coefficient of the three-dimensional scene camera is monitored in real time, and the hierarchy table is instantly queried based on the coefficient, and the most suitable point cloud detail level is dynamically switched to render. After the interactive operation is completed, the system returns to the initial full-detail point cloud display state.
[0029] Preferably, in the step 2, the point cloud is downsampled and divided into levels by combining the KD-Tree data structure and the octree, so as to effectively retain the structural features and details of the point cloud.
[0030] Preferably, in step three, dynamic updating of multi-scale details is achieved by inputting the number of point clouds and camera coefficients to establish a coefficient hierarchy table, so that the point cloud detail display can be adjusted according to different observation distances.
[0031] Preferably, during the point cloud rendering process, the display quantity and detail level of the point cloud can be automatically adjusted according to different zoom scales to improve rendering efficiency and quality.
[0032] Compared with the prior art, the present invention has the following beneficial effects:
[0033] 1. The present invention establishes a coefficient hierarchy table according to the number of point clouds and camera coefficients, and uses KD-Tree and octree to perform point cloud downsampling and hierarchical division. It can automatically adjust the display quantity and detail level of point clouds according to different zoom scales, thereby significantly improving rendering efficiency and meeting the needs of real-time rendering.
[0034] 2. The present invention adopts a combination of KD-Tree and octree to perform point cloud downsampling and hierarchical division, which better preserves the structural characteristics and details of the point cloud, avoids the loss of important information when processing point clouds with rich details, and ensures high-quality rendering effects.
[0035] 3. The present invention layers the point cloud by inputting the number of point clouds and camera coefficients to establish a coefficient hierarchy table, thereby simplifying the operation process and improving the operating efficiency and continuity of the system.
[0036] 4. The present invention is suitable for real-time rendering and display of large-scale point clouds in multiple fields such as 3D modeling, virtual reality, autonomous driving, and high-precision maps, and can meet the needs of large-scale point cloud rendering in different scenarios.
[0037] 5. The present invention presents high-quality point cloud rendering effects while ensuring real-time rendering, providing users with better visual and interactive experiences in different scenarios. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:
[0039] Figure 1 A schematic diagram of the process of the present invention;
[0040] Figure 2 Schematic diagram of the point cloud level downsampling process of the present invention. DETAILED DESCRIPTION
[0041] The following is a description of the implementation of the present invention by means of specific embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification.
[0042] See also Figure 1 to Figure 2 . It should be noted that the structures, proportions, sizes, etc. illustrated in the drawings in this specification are only used to match the contents disclosed in the specification for people familiar with this technology to understand and read, and are not used to limit the limiting conditions for the implementation of the present invention, so they have no substantial technical significance. Any modification of the structure, change in the proportional relationship, or adjustment of the size should still fall within the scope of the technical content disclosed by the present invention without affecting the effects and purposes that can be achieved by the present invention. At the same time, the terms such as "upper", "lower", "left", "right", "middle" and "one" quoted in this specification are only for the convenience of description, and are not used to limit the scope of the implementation of the present invention. Changes or adjustments in their relative relationships should also be regarded as the scope of the implementation of the present invention without substantially changing the technical content.
[0043] Embodiment 1, a large-scale point cloud rendering system based on a multi-scale detail model, comprising:
[0044] The point cloud reading module is used to read the point cloud file and determine the number of scales to be divided according to the number of points in the point cloud. The calculation formula of the number of scales num of the divided point cloud is: num=total / min, where total is the number of points in the point cloud and min is the minimum number of points to be displayed;
[0045] If the total number of points in the point cloud is 1,000,000 and the minimum number of points to be displayed is 200,000, then the number of point cloud scales num is 5;
[0046] Data combination and segmentation module, including:
[0047] KD-Tree data structure unit, used to efficiently combine point cloud data to perform nearest neighbor search of points;
[0048] The octree segmentation unit is used to segment the point cloud and iteratively divide the point cloud into num point clouds. The specific process is as follows:
[0049] The point cloud is stored in the KD-Tree data structure for subsequent representative point calculation;
[0050] Select a scale and use the octree structure to iteratively divide the point cloud until the number of cells is the same as the number of points corresponding to the scale, such as 400,000.
[0051] Use the DBSCAN algorithm to calculate the density points of the point cloud in the cell. During the calculation process, the nearest neighbor search is performed on each point to find the point with the highest density as the representative of the cell;
[0052] The coefficient hierarchy table establishment module is used to establish a system hierarchy table according to the 3D scene zoom coefficient. For example, the zoom coefficient is divided into 5 intervals according to the number of scales in the range of 0-500, and each interval corresponds to a scale. When the zoom coefficient exceeds 500, the lowest level detail point cloud is used for rendering and display;
[0053] The real-time monitoring and rendering module is used to monitor the orthogonal projection coefficients of the three-dimensional scene camera in real time when the user performs interactive operations such as zooming and panning, and to instantly query the hierarchy table based on this, dynamically switch to the most suitable point cloud detail level for rendering, and after the interactive operation is completed, the system immediately returns to the initial full-detail point cloud display state.
[0054] In the data combination and segmentation module, the point cloud is downsampled and divided into levels by combining the KD-Tree data structure and the octree, which can better preserve the structural features and details of the point cloud.
[0055] In the coefficient hierarchy table establishment module, dynamic updating of multi-scale details is achieved by inputting the number of point clouds and camera coefficients to establish a coefficient hierarchy table, so as to meet the requirements of point cloud detail display at different observation distances.
[0056] Embodiment 2, a large-scale point cloud rendering method based on a multi-scale detail model, comprises the following steps:
[0057] Step 1: Point cloud reading and scale determination, read the point cloud file, determine the number of scales divided according to the number of points in the point cloud, the point cloud scale number num calculation formula is: num = total / min, where total is the number of points in the point cloud, and min is the minimum number of points to be displayed;
[0058] Step 2: Data combination and segmentation, including:
[0059] Use KD-Tree data structure to efficiently combine point cloud data for nearest neighbor search of points;
[0060] Use the octree to segment the point cloud and iteratively divide the point cloud into num point clouds, including:
[0061] The point cloud is stored in the KD-Tree data structure for subsequent representative point calculation;
[0062] Select a scale and iteratively divide the point cloud using an octree structure until the number of cells divided is the same as the number of points corresponding to the scale;
[0063] Use the DBSCAN algorithm to calculate the density points of the point cloud in the cell, and find the point with the highest density as the representative of the cell by performing a nearest neighbor search on each point;
[0064] Step 3: Establish a coefficient hierarchy table. Establish a system hierarchy table based on the 3D scene scaling coefficient. The scaling coefficient is divided according to the scale quantity in the range of 0-500. When the scaling coefficient exceeds 500, the lowest level detail point cloud is used for rendering and display.
[0065] Step 4: Real-time monitoring and rendering. When the user performs interactive operations such as zooming and panning, the orthogonal projection coefficient of the three-dimensional scene camera is monitored in real time, and the hierarchy table is instantly queried based on the coefficient, and the most suitable point cloud detail level is dynamically switched to render. After the interactive operation is completed, the system returns to the initial full-detail point cloud display state.
[0066] In the step 2, the point cloud is downsampled and divided into levels by combining the KD-Tree data structure and the octree, so as to effectively retain the structural features and details of the point cloud.
[0067] In the step three, the dynamic update of multi-scale details is achieved by inputting the number of point clouds and camera coefficients to establish a coefficient hierarchy table, so that the point cloud detail display can be adjusted according to different observation distances.
[0068] During the point cloud rendering process, the display quantity and detail level of the point cloud can be automatically adjusted according to different zoom scales to improve rendering efficiency and quality.
[0069] Embodiment 3, a large-scale point cloud rendering method based on a multi-scale detail model, the specific implementation steps are as follows:
[0070] Point cloud reading and scale determination:
[0071] Read the point cloud file and get the number of points in the point cloud.
[0072] Assuming that the number of points in the point cloud is 800,000 and the minimum number of points to be displayed is 100,000, the number of scales to be divided is 8.
[0073] Data combination and segmentation:
[0074] The KD-Tree data structure is used to combine point cloud data to facilitate the nearest neighbor search of points.
[0075] Use octree to segment the point cloud:
[0076] The point cloud is stored in the KD-Tree data structure.
[0077] Select scale 3 and use the octree structure to iteratively divide the point cloud until the number of cells is the same as the number of points corresponding to the scale (such as 200,000).
[0078] The DBSCAN algorithm is used to calculate the density points of the point cloud in the cell, and the point with the highest density is found as the representative of the cell through the nearest neighbor search.
[0079] Coefficient hierarchy table creation:
[0080] A system hierarchy table is established according to the zoom factor of the 3D scene. For example, the zoom factor is divided into 8 intervals according to the number of scales in the range of 0-500, and each interval corresponds to a scale. When the zoom factor exceeds 500, the lowest level of detail point cloud is used for rendering and display.
[0081] Real-time monitoring and rendering:
[0082] When the user performs a zoom operation, the orthogonal projection coefficients of the 3D scene camera are monitored in real time.
[0083] Assuming that the scaling factor corresponding to the current orthogonal projection coefficient is in interval 4, the system instantly queries the level table and dynamically switches to the corresponding point cloud detail level for rendering.
[0084] After the interactive operation is completed, the system returns to the initial full-detail point cloud display state.
[0085] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by a person of ordinary skill in the art without departing from the spirit and technical concept disclosed by the present invention shall still be covered by the claims of the present invention.
Claims
1. A large-scale point cloud rendering system based on a multi-scale detail model, characterized in that: include: The point cloud reading module is used to read the point cloud file and determine the number of scales to be divided according to the number of points in the point cloud. The calculation formula of the number of scales num of the divided point cloud is: num=total / min, where total is the number of points in the point cloud and min is the minimum number of points to be displayed; Data combination and segmentation module, including: KD-Tree data structure unit, used to efficiently combine point cloud data to perform nearest neighbor search of points; The octree segmentation unit is used to segment the point cloud and iteratively divide the point cloud into num point clouds. The specific process is as follows: The point cloud is stored in the KD-Tree data structure for subsequent representative point calculation; Select a scale and iteratively divide the point cloud using an octree structure until the number of cells divided is the same as the number of points corresponding to the scale; Use the DBSCAN algorithm to calculate the density points of the point cloud in the cell. During the calculation process, the nearest neighbor search is performed on each point to find the point with the highest density as the representative of the cell; A coefficient hierarchy table establishment module is used to establish a system hierarchy table according to the three-dimensional scene scaling coefficient, where the scaling coefficient is divided according to the scale number in the range of 0-500. When the scaling coefficient exceeds 500, the lowest level detail point cloud is used for rendering and display; The real-time monitoring and rendering module is used to monitor the orthogonal projection coefficients of the three-dimensional scene camera in real time when the user performs interactive operations such as zooming and panning, and to instantly query the hierarchy table based on this, dynamically switch to the most suitable point cloud detail level for rendering, and after the interactive operation is completed, the system immediately returns to the initial full-detail point cloud display state.
2. The large-scale point cloud rendering system based on a multi-scale detail model according to claim 1, characterized in that: In the data combination and segmentation module, the point cloud is downsampled and divided into levels by combining the KD-Tree data structure and the octree, which can better preserve the structural features and details of the point cloud.
3. The large-scale point cloud rendering system based on a multi-scale detail model according to claim 1, characterized in that: In the coefficient hierarchy table establishment module, dynamic updating of multi-scale details is achieved by inputting the number of point clouds and camera coefficients to establish a coefficient hierarchy table, so as to meet the requirements of point cloud detail display at different observation distances.
4. A large-scale point cloud rendering method based on a multi-scale detail model, characterized in that: The following steps are involved: Step 1: Point cloud reading and scale determination, read the point cloud file, determine the number of scales divided according to the number of points in the point cloud, the point cloud scale number num calculation formula is: num = total / min, where total is the number of points in the point cloud, and min is the minimum number of points to be displayed; Step 2: Data combination and segmentation, including: Use KD-Tree data structure to efficiently combine point cloud data for nearest neighbor search of points; Use the octree to segment the point cloud and iteratively divide the point cloud into num point clouds, including: The point cloud is stored in the KD-Tree data structure for subsequent representative point calculation; Select a scale and iteratively divide the point cloud using an octree structure until the number of cells divided is the same as the number of points corresponding to the scale; Use the DBSCAN algorithm to calculate the density points of the point cloud in the cell, and find the point with the highest density as the representative of the cell by performing a nearest neighbor search on each point; Step 3: Establish a coefficient hierarchy table. Establish a system hierarchy table based on the 3D scene scaling coefficient. The scaling coefficient is divided according to the scale quantity in the range of 0-500. When the scaling coefficient exceeds 500, the lowest level detail point cloud is used for rendering and display. Step 4: Real-time monitoring and rendering. When the user performs interactive operations such as zooming and panning, the orthogonal projection coefficient of the three-dimensional scene camera is monitored in real time, and the hierarchy table is instantly queried based on the coefficient, and the most suitable point cloud detail level is dynamically switched to render. After the interactive operation is completed, the system returns to the initial full-detail point cloud display state.
5. The large-scale point cloud rendering method based on a multi-scale detail model according to claim 4, characterized in that: In the step 2, the point cloud is downsampled and divided into levels by combining the KD-Tree data structure and the octree, so as to effectively retain the structural features and details of the point cloud.
6. The large-scale point cloud rendering method based on a multi-scale detail model according to claim 4, characterized in that: In the step three, the dynamic update of multi-scale details is achieved by inputting the number of point clouds and camera coefficients to establish a coefficient hierarchy table, so that the point cloud detail display can be adjusted according to different observation distances.
7. The large-scale point cloud rendering method based on a multi-scale detail model according to claim 4, characterized in that: During the point cloud rendering process, the display quantity and detail level of the point cloud can be automatically adjusted according to different zoom scales to improve rendering efficiency and quality.
Citation Information
Cited By
Access method and device of point cloud data, storage medium and electronic equipment
CN119271678A
Live-action three-dimensional geographic entity and space planning display system
CN120689537A
Dynamic and static combined multi-scale point cloud adaptive characterization method and system
CN120852764A
