Point cloud based 3D reconstruction method for operation scene of distribution line

Abstract

The invention discloses a point cloud based 3D reconstruction method for an operation scene of a distribution line. Operations including conditional filtering, de-sampling, outlier removing and segmenting are carried out on an initial point cloud to obtain point cloud of to-be-constructed parts, as a lightening arrester and a cross arm, of the distribution line, a SIFT 3D algorithm is used to extract key points, FPFH characteristics are used to construct describing vectors of the key points, coarse registering and improved ICP fine registering are used to complete registering of the point cloud, a complete 3D point cloud of the parts is obtained, an offline model library is established, the point cloud is collected in real time, the point cloud of the parts is registered with models of themodel library, 3D reconstruction is completed, curved surfaces are reconstructed via a Poisson surface reconstruction method, and a 3D model is obtained. According to the invention, semi-autonomous 3D reconstruction of the distribution line scene is realized, manual intervention is reduced, aimed registering serves as an important step of 3D reconstruction, SIFT 3D key point extraction and FPFH characteristic describing vectors are used, the quality of the key points is ensured, point pair relation is provided with weights, incorrect point pairs are eliminated, the registering speed is improved, and the 3D reconstruction efficiency is improved.

Description

technical field [0001] The invention relates to the field of environment perception of live working robots, in particular to a method for three-dimensional reconstruction of power distribution line operation scenes based on point clouds. Background technique [0002] With the development of robot technology, robots are playing an increasingly important role in various fields. Applying robot technology to the power industry to replace manual power maintenance and repair work can greatly improve the safety and efficiency of operations. [0003] The use of robots for live work generally has a teleoperation mode and an autonomous mode, and no matter which mode it is, it is necessary to perform 3D reconstruction of the operation scene. Strong human-computer interaction operation; the second is to enable the robot to have scene perception capabilities, and can perform autonomous obstacle avoidance and motion planning [0004] The 3D reconstruction based on point cloud mainly inc...

AI Technical Summary

Problems solved by technology

For coarse registration, it is usually necessary to extract the geometric features of the point cloud to prepare for the establishment of the corresponding relationship of the point cloud. The current popular coarse registration method based on RANSAC is relatively random, which affects the registration efficiency and reduces the modeling effort. speed

For fine registration, the ICP algorithm is generally used, but the traditional ICP algorithm uses all point sets as the point set to be registered, and does not exclude error point pairs, which not only affects the speed of registration, but also reduces the accuracy of registration

Images