Rapid scene modeling method based on point cloud three-dimensional sparse matrix linear storage structure

Abstract

The invention discloses a rapid scene modeling method based on a point cloud three-dimensional sparse matrix linear storage structure. The method comprises the steps of data preprocessing, linear index mechanism and scene detail level construction, and point cloud data storage and retrieval. According to the method, the addressing cost of a traditional spatial data structure is eliminated, the spatial complexity is reduced, efficient storage and rapid retrieval of large-scale point cloud data are achieved, and the proposed rapid scene modeling method has remarkable advantages in the aspects of data processing efficiency and rendering performance and is suitable for being popularized and applied. The method is suitable for scenes needing efficient spatial data processing and real-time rendering, such as wide-area-scale topographic data acquisition and analysis, and has a good application prospect.

Description

Technical Field

[0001] The present invention belongs to the technical field of large-scale three-dimensional space point cloud processing, and in particular relates to a fast scene modeling method based on a point cloud three-dimensional sparse matrix linear storage structure. Background Art

[0002] Large-scale 3D terrain scenes are an extremely important component of 3D virtual environments. The rapid and efficient acquisition of scene data provides a basis for spatial analysis and deformation monitoring. In spatial analysis and deformation monitoring, especially for wide-scale slopes, due to the large area and complex geological structure, the collected terrain data sets are large and the surface features are complex and diverse. The large scale of point cloud data sets will inevitably have an adverse effect on subsequent processing, storage, transmission and display. Geomorphic data is basic data, and a series of technologies have been developed around its acquisition needs.

[0003] With the rapid development of hardware technology, the detailed information of the target scene is becoming more and more refined. Whether a scene is expressed by a large-scale spatial point cloud or a voxel scene grid, the final amount of data has shown an explosive growth. The growth rate of data volume has far exceeded the development speed of computer hardware.

[0004] The existing point cloud 3D scene modeling technology is not efficient in the use of hardware resources, especially in the processing, storage and display of large-scale point cloud data, which has high hardware requirements, while the development speed of hardware technology has not kept up with the growth rate of data volume. Therefore, it is urgent to find an efficient, accurate and hardware-saving large-scale point cloud storage and fast scene modeling method. Summary of the invention

[0005] In order to solve the technical problems raised in the background technology, the purpose of the present invention is to provide a fast scene modeling method based on a linear storage structure of a point cloud three-dimensional sparse matrix.

[0006] The object of the present invention is achieved by comprising the following steps: S100, data preprocessing: reordering the original disordered point cloud to form an ordered data set; S200, constructing a linear indexing mechanism and scene detail level: constructing a linear indexing mechanism for ordered data using a linear octree index and a spatial data adjacent sorting mechanism, and establishing a scene detail level; S300, point cloud data storage and retrieval: using the sparse octree data structure, point cloud data is stored and retrieved by dividing the space into layers.

[0007] Preferably, step S100 reorders the original unordered point cloud, including the following steps: S101, spatially sorting the original point cloud data, using a pointer to point to the starting point of the child node array, and then using an offset for addressing to construct a node-coupled sparse octree to form an ordered data set; S102. Use a flag variable to mark the child node status.

[0008] Preferably, step S200 uses the linear octree index and the spatial data adjacent sorting mechanism to construct a linear index mechanism for the ordered data, specifically including the following steps: S201, linearizing the sparse octree structure to complete the scheduling of nodes and data contained in the tree structure; S202, constructing a vertex index pool and a dynamic linked list based on integer offsets; reducing storage space requirements, achieving fast data retrieval and scheduling, and meeting the requirements of real-time rendering at a high frame rate, it is necessary to construct a vertex index pool and a dynamic linked list based on integer offsets by using a compact structure and an efficient indexing mechanism; S203. Sort and manage data using index linked lists and Morton coding.

[0009] Preferably, step S201 constructs an information domain for each sparse octree child node to control the expansion brought about by the increase in the depth of the tree; the information domain of the sparse octree child node includes a flag variable for indicating the node state, a variable of RGB type for representing the brightness of the current node, a point count variable for specifying the total number of points in the current child node branch, and finally a joint variable of the memory is used to encode a pointer to the child node, data, and the offset of the node in the memory.

[0010] Preferably, during the construction process of step S202, the 8-byte pointer is divided into two 4-byte variable structures, which are used to point to the offset of the next point in the file and to point to the end of the index chain respectively.

[0011] Preferably, the index linked list in step S203 includes: In the index linked list, each linked list node records the location information of a point cloud data; Retrieve point cloud data by sequentially traversing the linked list nodes; Use hash table to optimize the access speed of linked list.

[0012] Preferably, step S203 specifically includes the following steps: S2031. Determine the content of the node descriptor and the continuous storage rule of the child node descriptor: in the organization of data, the head node uses a single descriptor, and the child node uses an array to uniformly manage the descendant nodes; in the file layout of the node descriptor, the nodes are stored in a continuous manner according to the specific implementation, wherein blanks are used to represent the hierarchical structure of the node coupling array, which does not actually exist; S2032. Determine the traversal rules of the descriptor and establish an index tree: the information of the linear sparse octree is recorded in the file header, and the node descriptor is recorded in the file in the traversal order to describe the node; according to the actual computing power of the system, a flexible memory window is used to record the corresponding node table in the node search; the linear sparse octree is regarded as an index tree for fast scheduling of data sets; in addition, based on the built-in level of detail (LOD) information, LOD data access of different granularities can be achieved; After the index tree is established, a corresponding management mechanism needs to be established for the data in case the LOD detail level of the index tree cannot meet the corresponding presentation requirements, or the corresponding data needs to be quickly searched for operation; in the present invention, the data is stored in a separate file sequence; S2033. Establish a data management mechanism: Use Morton code to sort point cloud data to reduce the potential frequency of data exchange, and group and index point cloud data; the spatial proximity of Morton code can be used to improve data retrieval efficiency, making full use of the spatial proximity of data, bringing its own spatial LOD detail layer, and having significant performance advantages; using Morton code to sort point cloud data in leaf nodes is transparent, and access relies on index lists; after performing large-scale editing operations and causing fragmentation, the point cloud data in the leaf nodes are reorganized to ensure high-speed operation of the system.

[0013] Among them, due to the existence of the index link access mechanism, there is no mandatory requirement for the amount of data in the node, and the spatial proximity is simply used; the Morton encoding of the three-dimensional leaf node matrix is ​​similar to the two-dimensional case.

[0014] Preferably, step S200 establishes the detail level of the scene by: embedding detail level information in the index tree to support detail level data access of different granularities; when the data of the index leaf node cannot meet the detail level rendering requirements, retrieving the original data in the external memory for display; The detail level information specifically includes: recording the detail level of the node in each node descriptor; dynamically adjusting the rendering detail level according to the viewpoint distance; and using a lower detail level for areas far from the viewpoint to reduce the amount of calculation; Retrieving the original data in the external memory specifically includes: retrieving high-resolution data in the external memory when the detail level cannot meet the rendering requirements; preprocessing the high-resolution data to ensure its compatibility with existing data; and dynamically loading and unloading the high-resolution data according to the rendering requirements.

[0015] Preferably, step S300 specifically includes the following steps: Use sparse octree to perform spatial hierarchical partitioning of point cloud data; Record node information in the node descriptor and store it in traversal order; Multi-level detail management and fast data retrieval through the node hierarchy.

[0016] Preferably, the spatial hierarchical division comprises the following steps: Based on the sparse octree structure, the information domain of the child nodes is established; Use page mechanism or bucket mechanism to store leaf node data; Index non-leaf nodes and locate leaf nodes; The loading and unloading of leaf nodes are dynamically adjusted according to the data access frequency.

[0017] Preferably, step S300 uses ray casting and three-dimensional Brensenham algorithm to perform data retrieval during point cloud data storage and retrieval; and establishes a data management and scheduling mechanism.

[0018] Preferably, ray casting and three-dimensional Brensenham algorithm are used to perform data retrieval, specifically comprising the following steps: S3011, using the ray casting algorithm to determine the intersection of the ray and the octree node: first recursively test the parameters, such as the distance to be tracked from the nearest viewpoint, the three-dimensional depth of the current test, and then use the ray casting algorithm to perform the ray intersection test. During the test, the ray cube (ray bounding box) test is performed from the top of the octree. If there is an intersection, recursively go deeper until the middle layer cube is found or the leaf node is reached; S3012. Use the three-dimensional Brensenham algorithm to sample and calculate the stepping of rays in three-dimensional space: through the ray equation, based on the coordinate range of the test area and the step length in each direction, quickly calculate the grid; for the application of subsequent algorithms, each grid is finally represented by its center point; by using the coordinates of the center point, the insertion test is performed in the order from near to far in the subsequent links, and finally the nearest intersection is obtained.

[0019] The 3D Brensenham algorithm converts the process of ray intersection test into point insertion test. The algorithm mainly includes an external exposure interface and the core process of given point insertion test.

[0020] Preferably, in step S3011, the nodes in the parent node are tested, and the direction of the recursive depth is determined according to the test result.

[0021] The recursive traversal algorithm mainly includes the comparison of the LOD of the intermediate points and the function of returning the point cloud data from the leaf nodes. According to the intersection test of the child nodes, the algorithm searches for the branch with the closest intersection for deep recursion.

[0022] Preferably, in establishing a data management and scheduling mechanism, point cloud data is dynamically scheduled based on the current visible range and user needs; an LRU (least recently used) mechanism is used to track and control the life cycle of nodes; memory usage is optimized to improve the response speed of data processing and rendering.

[0023] Compared with the prior art, the present invention has the following technical effects: 1. The method of the present invention eliminates the addressing cost of the traditional spatial data structure, reduces the spatial complexity, and realizes the efficient storage and rapid retrieval of large-scale point cloud data. The proposed rapid scene modeling method has significant advantages in data processing efficiency and rendering performance. It is suitable for scenes that require efficient spatial data processing and real-time rendering, such as wide-area scale terrain data collection and analysis, and has good application prospects. 3. The present invention uses the adjacency of spatial data to sort and manage, effectively reducing the time complexity of data processing and retrieval; 4. The present invention adopts a three-dimensional sparse matrix storage algorithm, combined with information compression coding, which significantly reduces the storage requirements of point cloud data; 5. The present invention adopts a flexible mechanism to allow the management and processing of large-scale data under limited memory; real-time scheduling of data according to the current visible range avoids the need to load all data into memory; 6. The present invention realizes flexible data scheduling, more efficient point cloud rendering and calculation under different CPU environments, especially in the application of terrain modeling and deformation monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a schematic diagram of the process of the present invention; Figure 2 It is a node-coupled sparse quadtree; Figure 3 Vertex index pool and index list for leaf nodes; Figure 4 It is the layout structure of the node descriptor; Figure 5 The file structure of the index tree; Figure 6 It is the Morton code of 8×8 two-dimensional matrix; Figure 7 Schematic diagram of ray casting sampling for a plane mesh; Figure 8 This is a rendering effect display. DETAILED DESCRIPTION

[0025] The present invention is further described below in conjunction with the embodiments and drawings, but the present invention is not limited in any way. Any changes or substitutions made based on the teachings of the present invention belong to the protection scope of the present invention.

[0026] Example 1 As attached Figure 1 As shown, the fast scene modeling method based on the linear storage structure of the point cloud three-dimensional sparse matrix in this embodiment includes the following steps: S100, data preprocessing: reordering the original disordered point cloud to form an ordered data set; S200, constructing a linear indexing mechanism and scene detail level: constructing a linear indexing mechanism for ordered data using a linear octree index and a spatial data adjacent sorting mechanism, and establishing a scene detail level; S300, point cloud data storage and retrieval: using the sparse octree data structure, point cloud data is stored and retrieved by dividing the space into layers.

[0027] Example 2 The fast scene modeling method based on the linear storage structure of the point cloud three-dimensional sparse matrix in this embodiment is based on the embodiment 1, and the step S100 reorders the original disordered point cloud, including the following steps: S101. Sort the original point cloud data in space, as shown in the attached Figure 2 As shown, considering the sparsity of the matrix and subsequent operations, a node-coupled sparse octree is constructed by using a pointer to point to the starting point of the child node array and then using an offset for addressing to form an ordered data set; S102. Use a flag variable to mark the state of the child node. In different operating environments or platforms, it is necessary to dynamically schedule the nodes according to actual resources. Use the flag variable to promote the traversal, dynamic loading and unloading of data in the sparse octree.

[0028] Example 3 The fast scene modeling method based on the linear storage structure of the point cloud three-dimensional sparse matrix in this embodiment is based on the embodiment 2. The step S200 uses the linear octree index and the spatial data adjacent sorting mechanism to construct a linear index mechanism for the ordered data, which specifically includes the following steps: S201, linearizing the sparse octree structure to complete the scheduling of nodes and data contained in the tree structure; S202, constructing a vertex index pool and a dynamic linked list based on integer offsets; S203. Sort and manage data using index linked lists and Morton coding.

[0029] Example 4 The fast scene modeling method based on the linear storage structure of the three-dimensional sparse matrix of the point cloud in this embodiment is based on the embodiment 3. Step S201 constructs an information domain for each sparse octree child node to control the expansion brought about by the increase of the depth of the tree; the information domain of the sparse octree child node includes a flag variable for indicating the node state, a variable "color" of RGB type for representing the brightness of the current node, a point count variable for specifying the total number of points in the current child node branch, and finally a joint variable of the memory is used to encode a pointer to the child node, data and the offset of the node in the memory.

[0030] Example 5 The fast scene modeling method based on the linear storage structure of the point cloud three-dimensional sparse matrix in this embodiment is based on Example 4. During the construction process of step S202, the 8-byte pointer is divided into two 4-byte variable structures, which are used to point to the offset of the next point in the file and the end of the index chain respectively.

[0031] Example 6 The fast scene modeling method based on the linear storage structure of the point cloud three-dimensional sparse matrix in this embodiment is based on the embodiment 5. The index linked list in step S203 includes: In the index linked list, each linked list node records the location information of a point cloud data; Retrieve point cloud data by sequentially traversing the linked list nodes; Use hash table to optimize the access speed of linked list.

[0032] Example 7 As attached Figure 4 , Attachment Figure 5 , Attachment Figure 6 As shown, the fast scene modeling method based on the linear storage structure of the point cloud three-dimensional sparse matrix in this embodiment is based on Example 6, and step S203 specifically includes the following steps: S2031. Determine the content of the node descriptor and the continuous storage rule of the child node descriptor: in the organization of data, the head node uses a single descriptor, and the child node uses an array to uniformly manage the descendant nodes; as shown in FIG. 4 , in the file layout of the node descriptor, the nodes are stored in a continuous manner according to the specific implementation, wherein blanks are used to represent the hierarchical structure of the node coupling array, which does not actually exist; S2032. Determine the traversal rules of the descriptor and establish an index tree: the information of the linear sparse octree is recorded in the file header, and the node descriptors are recorded in the file in the traversal order to describe the nodes; according to the actual computing power of the system, a flexible memory window is used to record the corresponding node table in the node search; the linear sparse octree is regarded as an index tree for fast scheduling of data sets; the index tree structure is as shown in the attached figure. Figure 5 As shown; S2033. Establish data management mechanism: as attached Figure 6 As shown, Morton codes are used to sort point cloud data to reduce the potential exchange frequency of data, and point cloud data are grouped and indexed; the spatial proximity of Morton codes can be used to improve data retrieval efficiency. Figure 6 In the process, the point cloud data contained in the leaf nodes are Morton encoded, which makes full use of the spatial proximity of the data and brings its own spatial LOD detail layer, which has significant performance advantages; the use of Morton encoding to sort the point cloud data in the leaf nodes is transparent, and the access relies on the index list; after performing large-scale editing operations and causing fragmentation, the point cloud data in the leaf nodes are reorganized to ensure the high-speed operation of the system.

[0033] Example 8 The fast scene modeling method based on the linear storage structure of the point cloud three-dimensional sparse matrix in this embodiment is based on the embodiment 7. The step S200 establishes the detail level of the scene by: embedding the detail level information in the index tree to support the access of detail level data of different granularities; when the data of the index leaf node cannot meet the detail level rendering requirements, retrieving the original data in the external memory for display; The detail level information specifically includes: recording the detail level of the node in each node descriptor; dynamically adjusting the rendering detail level according to the viewpoint distance; and using a lower detail level for areas far from the viewpoint to reduce the amount of calculation; Retrieving the original data in the external memory specifically includes: retrieving high-resolution data in the external memory when the detail level cannot meet the rendering requirements; preprocessing the high-resolution data to ensure its compatibility with existing data; and dynamically loading and unloading the high-resolution data according to the rendering requirements.

[0034] Example 9 The fast scene modeling method based on the linear storage structure of the point cloud three-dimensional sparse matrix in this embodiment is based on the embodiment 8, and the step S300 specifically includes the following steps: Use sparse octree to perform spatial hierarchical partitioning of point cloud data; Record node information in the node descriptor and store it in traversal order; Multi-level detail management and data retrieval are performed through the node hierarchy.

[0035] Example 10 The fast scene modeling method based on the linear storage structure of the point cloud three-dimensional sparse matrix in this embodiment is based on the embodiment 9, and the spatial hierarchical division includes the following steps: Based on the sparse octree structure, the information domain of the child nodes is established; Use page mechanism or bucket mechanism to store leaf node data; Index non-leaf nodes and locate leaf nodes; The loading and unloading of leaf nodes are dynamically adjusted according to the data access frequency.

[0036] Embodiment 11 The fast scene modeling method based on the linear storage structure of the point cloud three-dimensional sparse matrix in this embodiment is based on the embodiment 10. In step S300, ray casting and three-dimensional Brensenham algorithm are used to retrieve data during the point cloud data storage and retrieval process; a data management and scheduling mechanism is established.

[0037] Example 12 As attached Figure 7 As shown, the fast scene modeling method based on the linear storage structure of the point cloud three-dimensional sparse matrix in this embodiment is based on Example 11, and ray casting and three-dimensional Brensenham algorithm are used for data retrieval, which specifically includes the following steps: S3011, using the ray casting algorithm to determine the intersection of the ray and the octree node: first set the recursive test parameters, then use the ray casting algorithm to perform the ray intersection test, and during the test, perform the ray cube test from the top of the octree. If there is an intersection, recursively go deeper until the middle layer cube is found or the leaf node is reached; S3012. Use the three-dimensional Brensenham algorithm to sample and calculate the stepping of rays in three-dimensional space: calculate the grid through the ray equation based on the coordinate range of the test area and the step length in each direction; for the application of subsequent algorithms, each grid is finally represented by its center point; by using the coordinates of the center point, the insertion test is performed in the order from near to far in the subsequent links, and finally the nearest intersection is obtained.

[0038] Embodiment 13 The fast scene modeling method based on the linear storage structure of the point cloud three-dimensional sparse matrix in this embodiment is based on Embodiment 12, and in step S3011, the nodes in the parent node are tested, and the direction of the recursive depth is determined according to the test results.

[0039] Embodiment 14 The fast scene modeling method based on the linear storage structure of the point cloud three-dimensional sparse matrix in this embodiment is based on Example 13. In establishing a data management and scheduling mechanism, point cloud data is dynamically scheduled based on the current visible range and user needs; an LRU mechanism is used to track and control the life cycle of nodes; and memory usage is optimized to improve the response speed of data processing and rendering.

[0040] Embodiment 14 Using the multi-viewpoint 3D laser point cloud data collected in a certain area, the effect of rendering the scene according to the method of Example 13 is shown in the attached figure. Figure 8 shown.

Claims

1. A fast scene modeling method based on point cloud three-dimensional sparse matrix linear storage structure, characterized by The following steps are involved: S100, data preprocessing: reordering the original disordered point cloud to form an ordered data set; S200, constructing a linear indexing mechanism and scene detail level: constructing a linear indexing mechanism for ordered data using a linear octree index and a spatial data adjacent sorting mechanism, and establishing a scene detail level; S300, point cloud data storage and retrieval: using the sparse octree data structure, point cloud data is stored and retrieved by dividing the space into layers.

2. The fast scene modeling method based on the point cloud three-dimensional sparse matrix linear storage structure according to claim 1 is characterized in that Step S100 reorders the original unordered point cloud, including the following steps: S101, spatially sorting the original point cloud data, using a pointer to point to the starting point of the child node array, and then using an offset for addressing to construct a node-coupled sparse octree to form an ordered data set; S102. Use a flag variable to mark the child node status.

3. The fast scene modeling method based on point cloud three-dimensional sparse matrix linear storage structure according to claim 1 is characterized in that Step S200 uses the linear octree index and the spatial data adjacent sorting mechanism to construct a linear index mechanism for the ordered data, which specifically includes the following steps: S201, linearizing the sparse octree structure to complete the scheduling of nodes and data contained in the tree structure; S202, constructing a vertex index pool and a dynamic linked list based on integer offsets; S203. Sort and manage data using index linked lists and Morton coding.

4. The fast scene modeling method based on point cloud three-dimensional sparse matrix linear storage structure according to claim 3 is characterized in that Step S201 constructs an information domain for each sparse octree child node to control the expansion as the depth of the tree increases; The information domain of a sparse octree child node includes a flag variable to indicate the node status, an RGB type variable to represent the brightness of the current node, a point count variable to specify the total number of points in the current child node branch, and finally a memory union variable is used to encode pointers to child nodes, data, and node offsets in memory.

5. The rapid scene modeling method based on point cloud three-dimensional sparse matrix linear storage structure according to claim 3 is characterized in that During the construction process of step S202, the 8-byte pointer is divided into two 4-byte variable structures, which are used to point to the offset of the next point in the file and the end of the index chain respectively.

6. The fast scene modeling method based on point cloud three-dimensional sparse matrix linear storage structure according to claim 3 is characterized in that The index linked list in step S203 includes: In the index linked list, each linked list node records the location information of a point cloud data; Retrieve point cloud data by sequentially traversing the linked list nodes; Use hash tables to optimize linked list access speed.

7. The fast scene modeling method based on point cloud three-dimensional sparse matrix linear storage structure according to claim 3 is characterized in that Step S203 specifically includes the following steps: S2031. Determine the content of the node descriptor and the continuous storage rule of the child node descriptor: in the organization of data, the head node uses a single descriptor, and the child node uses an array to uniformly manage the descendant nodes; in the file layout of the node descriptor, the nodes are stored in a continuous manner according to the specific implementation, wherein blanks are used to represent the hierarchical structure of the node coupling array, which does not actually exist; S2032, determine the traversal rules of the descriptor and establish an index tree: the information of the linear sparse octree is recorded in the file header, and the node descriptor is recorded in the file in the traversal order to describe the node; according to the actual computing power of the system, a flexible memory window is used to record the corresponding node table in the node search; the linear sparse octree is regarded as an index tree for fast scheduling of the data set; S2033. Establish a data management mechanism: use Morton code to sort the point cloud data, and group and index the point cloud data; after performing large-scale editing operations and causing fragmentation, reorganize the point cloud data in the leaf nodes.

8. The fast scene modeling method based on point cloud three-dimensional sparse matrix linear storage structure according to claim 1 is characterized in that The step S200 of establishing the detail level of the scene specifically includes: embedding detail level information in the index tree to support detail level data access of different granularities; when the data of the index leaf node cannot meet the detail level rendering requirements, retrieving the original data in the external memory for display; The detail level information specifically includes: recording the detail level of the node in each node descriptor; dynamically adjusting the rendering detail level according to the viewpoint distance; and using a lower detail level for areas far from the viewpoint to reduce the amount of calculation; Retrieving the original data in the external memory specifically includes: retrieving high-resolution data in the external memory when the detail level cannot meet the rendering requirements; preprocessing the high-resolution data to ensure its compatibility with existing data; and dynamically loading and unloading the high-resolution data according to the rendering requirements.

9. The fast scene modeling method based on point cloud three-dimensional sparse matrix linear storage structure according to claim 1, characterized in that Step S300 specifically includes the following steps: Use sparse octree to perform spatial hierarchical partitioning of point cloud data; Record node information in the node descriptor and store it in traversal order; Multi-level detail management and data retrieval are performed through the node hierarchy.

10. The fast scene modeling method based on point cloud three-dimensional sparse matrix linear storage structure according to claim 9, characterized in that The spatial hierarchical division comprises the following steps: Based on the sparse octree structure, the information domain of the child nodes is established; Use page mechanism or bucket mechanism to store leaf node data; Index non-leaf nodes and locate leaf nodes; The loading and unloading of leaf nodes are dynamically adjusted according to the data access frequency.

11. The fast scene modeling method based on point cloud three-dimensional sparse matrix linear storage structure according to claim 1, characterized in that Step S300 uses ray casting and 3D Brensenham algorithm to retrieve data during point cloud data storage and retrieval; establishes data management and scheduling mechanism.

12. The fast scene modeling method based on point cloud three-dimensional sparse matrix linear storage structure according to claim 11, characterized in that Ray casting and 3D Brensenham algorithm are used for data retrieval, which includes the following steps: S3011, using the ray casting algorithm to determine the intersection of the ray and the octree node: first set the recursive test parameters, then use the ray casting algorithm to perform the ray intersection test, and during the test, perform the ray cube test from the top of the octree. If there is an intersection, recursively go deeper until the middle layer cube is found or the leaf node is reached; S3012. Use the three-dimensional Brensenham algorithm to sample and calculate the stepping of rays in three-dimensional space: calculate the grid through the ray equation based on the coordinate range of the test area and the step length in each direction; for the application of subsequent algorithms, each grid is finally represented by its center point; by using the coordinates of the center point, the insertion test is performed in the order from near to far in the subsequent links, and finally the nearest intersection is obtained.

13. The fast scene modeling method based on point cloud three-dimensional sparse matrix linear storage structure according to claim 12, characterized in that In step S3011, the nodes in the parent node are tested, and the direction of the recursive depth is determined according to the test result.

14. The fast scene modeling method based on point cloud three-dimensional sparse matrix linear storage structure according to claim 11, characterized in that In establishing data management and scheduling mechanisms, point cloud data is dynamically scheduled based on the current visible range and user needs; the LRU mechanism is used to track and control the life cycle of nodes; Optimize memory usage and improve the responsiveness of data processing and rendering.

Images