In situ 3D reconstruction is the basis for measuring and monitoring large-scale working environments such as construction sites and coal mining surfaces. The purpose of this paper is to address the difficulties of on-site 3D reconstruction in large-scale dynamic environments. Based on panoramic rotating light detection and ranging sensors, a dynamic environment 3D reconstruction method combining occupancy grid and point cloud segmentation is proposed.
The algorithm fully combines the information between point clouds to process the points in the region of interest, improving the processing speed of the occupancy grid method and the accuracy of dynamic object filtering. Furthermore, this paper also proposes an incremental long-short horizon segmentation and reconstruction workflow, which performs different levels of point cloud differential segmentation for various types of dynamic objects.
This paper designs detailed physical and simulation experiments in different environments to verify the advantages and reliability of the algorithm. The entire system showed satisfactory performance in experiments.
The proposed dynamic environment 3D reconstruction method can effectively complete the 3D reconstruction of complex work environments, and the proposed incremental long-short horizon segmentation and reconstruction workflow further improves the accuracy of dynamic object removal.
