Multi-plane identification method for three-dimensional point cloud

Abstract

The invention discloses a multi-plane identification method for a three-dimensional point cloud. The method is characterized by comprising the following steps: step 1, in the three-dimensional point cloud with multiple planes, the possibility that a plane exists nearby is evaluated for each point, so that each point obtains a reliability score; step 2, a space density tree of the whole point cloudis constructed through plane segmentation based on the reliability score of each point, the normal deviation and the plane deviation between points; step 3, the constructed space density tree is cut,so that points on different planes are gathered on different sub-trees respectively. In this way, the multi-plane recognition is achieved. According to the invention, the space density tree model suitable for plane segmentation is established for the three-dimensional point cloud. Multiple planes in the point cloud are naturally segmented under the condition that the number of specified planes isnot required. The method has strong distinguishing capacity especially for parallel planes at a very close distance relative to each other. The method is high in recognition efficiency and accurate in recognition effect. The method has wide applicability for various point clouds with planes.

Description

technical field [0001] The invention relates to the fields of computer vision three-dimensional perception and three-dimensional point cloud processing, in particular to a multi-plane recognition method for three-dimensional point clouds. Background technique [0002] With the development and popularization of 3D scanning technology, photogrammetry and depth sensors, we can easily obtain massive 3D point cloud information. Compared with 2D image information, 3D point cloud can more accurately and intuitively describe the geometric properties of objects. In fact, the geometric shapes of life and industrial products are composed of geometric primitives such as planes, spheres, cylinders and cones. It is an important part of the field of 3D point cloud processing to automatically and efficiently analyze the geometric primitives of parts from point cloud data with rich information. [0003] As the basic geometric primitive, the plane is the most common and characteristic type ...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the deficiencies in the prior art, the present invention provides a multi-plane recognition method for 3D point cloud, which solves the problem of inaccurate estimation of the existing 3D point cloud plane recognition method

Images