Multi-dimensional feature integrated building point cloud hierarchical clustering segmentation method

Abstract

The invention discloses a multi-dimensional feature integrated building point cloud hierarchical clustering segmentation method. The method comprises the steps of carrying out initial segmentation on mutually discontinuous point cloud data in building point cloud data; through utilization of a G-K (Gustafson-Kessel) clustering algorithm and through combination of the spectral characteristics of the point cloud data, carrying out first layer fine segmentation on the point cloud data; and carrying out second layer fine segmentation on obtained normal vector characteristics and curvature characteristics of the point cloud data, wherein two times of segmentation are repeated until demands are satisfied. According to the building point cloud hierarchical clustering segmentation method, the density information of the point cloud data is scanned through utilization of multi-dimensional laser; on the premise of no prior knowledge, the initial segmentation is carried out on a plurality of point cloud data blocks which are spaced at relatively long distance and are relatively dense; moreover, through combination of the spectral characteristics and geometrical characteristics of the point cloud data, the fine segmentation is carried out on the initially segmented point cloud data blocks, until each point cloud data block has a single geometrical characteristic and modeling can be carried out through utilization of a simple mathematical model. According to the method, the building point cloud data can be extracted from surroundings, the building point cloud can be decomposed into different planes, and a good foundation is laid for reconstructing a building.

Description

technical field [0001] The invention relates to a point cloud data segmentation algorithm, in particular to a building point cloud hierarchical clustering and segmentation method. Background technique [0002] Using 3D laser scanning technology to scan and measure buildings, the point cloud data of buildings can be obtained. In order to extract the physical features of buildings based on point cloud data, that is, the shape, position, size, length and area of ​​buildings or their components and parts For the description of key geometric quantities such as volume and volume, it is necessary to use the 3D point cloud data of the building to reconstruct the building model. Whether the model is established correctly or not directly affects whether the measurement of the building's physical characteristics is correct, so it is very important to use the scanned point cloud data reasonably to establish a building point cloud model. However, building point cloud models often contai...

AI Technical Summary

Problems solved by technology

However, various existing segmentation algorithms usually use a certain feature of the point cloud data or multiple features of the point cloud data to try to separate the point cloud data at one time. It is easy to cause under-segmentation or over-segmentation, and it is difficult to achieve correct segmentation of complex building point cloud data

Images