Local density-based octree point cloud preprocessing method

Abstract

The invention discloses a local density-based octree point cloud preprocessing method and the method comprises the steps: carrying out point cloud preprocessing based on an octree for an original data set, carrying out space division by a topological structure based on a density value octree, carrying out octree division of a subspace meeting a certain density value. The reconstruction efficiency is improved, meanwhile, waste of space resources is avoided, point cloud data is simplified and compressed after the topological structure is constructed, part of redundant data is reduced, finally, the point cloud data is inquired through radius neighborhood search, and basic work is made for subsequent reconstruction of the point cloud data.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to an octree point cloud preprocessing method based on local density. Background technique [0002] The point cloud data of the power equipment in the power plant has a large amount of original scattered point cloud data, the data processing speed is slow, and there is no topological connection between the point clouds, so the original data must be preprocessed and compressed; before point cloud compression, it is necessary to Select an appropriate point cloud topology for division. Space division can simplify the operation between points by establishing a data structure, and improve the efficiency and reconstruction effect of the reconstruction process. Therefore, it is necessary to select an appropriate data structure and algorithm to divide the space. Point cloud space There are three main division methods, K-D tree method, octree method and bounding box metho...

AI Technical Summary

Problems solved by technology

[0003] Through research, it can be found that when constructing a K-D tree, it is difficult to update the tree, the cost of building and pruning is very high, and the time complexity of subdividing the three-dimensional space is relatively high. Secondly, the K-D tree cannot be stored and accessed randomly. ; while the bounding box method will uniformly process the point cloud data of the sub-cube, but ignores the density difference of the point cloud, making the reconstruction result too smooth, unable to show the details of the object and the density difference, and the acquisition by the 3D scanner may produce A lot of noise or redundant points, so the bounding box is not suitable for point cloud space division

Images