Point cloud compression method and device, equipment, storage medium and program product

Abstract

The invention discloses a point cloud compression method and device, equipment, a storage medium and a program product, and belongs to the technical field of point cloud compression. The method comprises the following steps: classifying given point cloud data to obtain approximate ground point cloud, non-ground object point cloud and noise point cloud; performing hybrid tree structure-based coding compression on the approximate ground point cloud; performing predictive coding based on spatial rearrangement on the non-ground object point cloud; performing serialized compression based on Morton codes on the noise point cloud; and integrating the compressed data of each classification point cloud, and outputting a final compressed data stream. The embodiment of the invention can improve the compression gain.

Claims

1. A point cloud compression method, characterized in that: include: Classify the given point cloud data to obtain approximate ground point cloud, non-ground object point cloud and noise point cloud; Performing coding compression based on a hybrid tree structure on the approximate ground point cloud; Implementing predictive coding based on spatial rearrangement on the non-ground object point cloud; Performing Morton code-based serialization compression on the noise point cloud; Integrate the compressed data of each classified point cloud and output the final compressed data stream.

2. The point cloud compression method according to claim 1, characterized in that: The given point cloud data is classified to obtain approximate ground point cloud, non-ground object point cloud and noise point cloud, including: Mapping the point cloud data into a three-dimensional Boolean matrix through voxelization, adjusting the scale of the point cloud data based on a scale factor, and generating a compact spatial structure with a topological relationship; Constructing a KD tree spatial index based on the compact spatial structure, traversing from the root node in depth, calculating the number of neighbors of the node within a specified radius, classifying the points whose number of neighbors is lower than a threshold as noise point clouds, and outputting a non-noise point cloud; The non-noise point cloud is subjected to cloth simulation filtering processing, and approximate ground point cloud and non-ground object point cloud are separated according to the point cloud curvature and elevation distribution characteristics.

3. The point cloud compression method according to claim 1, characterized in that: The performing coding compression based on a hybrid tree structure on the approximate ground point cloud comprises: Generate an initial bounding box according to the spatial distribution of the approximate ground point cloud, and calculate its length, width and height; Based on the length, width and height of the initial bounding box, performing hybrid tree structure division on the approximate ground point cloud according to a preset division rule to generate a binary tree-quadtree-octree hybrid tree structure; Based on the binary tree-quadtree-octree hybrid tree structure, adjacent nodes or parent nodes are used to predict the current node to obtain a prediction result; The residuals of each prediction mode are calculated, the prediction mode corresponding to the residual with the minimum absolute value is selected, and the prediction result and the residual data are input into an arithmetic encoder to generate a first compressed bit stream.

4. The point cloud compression method according to claim 3, characterized in that: The method of performing hybrid tree structure division on the approximate ground point cloud based on the length, width and height of the initial bounding box according to a preset division rule to generate a binary tree-quadtree-octree hybrid tree structure includes: Define parameters K and M, where K represents the number of binary tree-quadtree mixed partitioning layers, and M represents the number of octree partitioning layers; For the first K layers, when the length, width and height of the initial bounding box are not equal, the approximate ground point cloud is divided into a binary tree along the axis of the maximum size; or, when there are two axes with equal sizes, the approximate ground point cloud is divided into a quadtree on the corresponding plane; For the latter M layers, octree partitioning is performed on the approximate ground point cloud to obtain a binary tree-quadtree-octree hybrid tree structure.

5. The point cloud compression method according to claim 1, characterized in that: The performing predictive coding based on spatial rearrangement on the non-ground object point cloud comprises: Based on the Euclidean clustering algorithm, the non-ground object point cloud is divided into multiple subsets by setting a distance threshold and adding an angle threshold; A directional bounding box algorithm is used to perform 3D packing on each of the subsets and mark the minimum coordinate value point. The traversal starts from the bounding box corresponding to the minimum coordinate value point, and each bounding box is translated to be adjacent to the previous bounding box and the translation amount is recorded to construct a compact non-ground object point cloud. The compact non-ground object point cloud is input into the TMC13 framework for lossless prediction coding and entropy coding to generate a second compressed bit stream.

6. The point cloud compression method according to claim 1, characterized in that: The performing of serialization compression based on Morton code on the noise point cloud comprises: Encoding and sorting the noise point cloud based on Morton code to obtain a noise point cloud sequence; The noise point cloud sequence is converted into binary Morton code to obtain a difference, and the residual value between adjacent points is recorded, and the residual value is compressed by run-length coding to obtain a third compressed bit stream.

7. A point cloud compression device, characterized in that: include: The classification module is used to classify the given point cloud data to obtain approximate ground point cloud, non-ground object point cloud and noise point cloud; A processing module, used for performing coding compression based on a hybrid tree structure on the approximate ground point cloud; A coding module, used for performing predictive coding based on spatial rearrangement on the non-ground object point cloud; A compression module, used for performing serialization compression based on Morton code on the noise point cloud; The integration module is used to integrate the compressed data of each classified point cloud and output the final compressed data stream.

8. An electronic device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the point cloud compression method according to any one of claims 1 to 6 when executing the computer program.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the point cloud compression method according to any one of claims 1 to 6.

10. A computer program product, characterized in that The computer program product comprises a computer program or a computer instruction, and when the computer program or the computer instruction is executed by a processor, the point cloud compression method according to any one of claims 1 to 6 is implemented.

Images