Extraction of 3D images using pitch-actuated 2D laser range finder for robotic vision

摘要

Volumetric mapping is essential for recognizing the 3D environment and consequentially for the decision making process for special-task mobile robots (such as stair climbing, inspection, or search and rescue). In this paper, we propose a fast surface detection algorithm based on point cloud data from a small pitch-actuated laser range finder mounted on a mobile manipulator arm. The developed algorithm is mainly composed of two steps: (i) gradient based classification; and (ii) Recursive Connected Component Labeling (RCCL) algorithm. The algorithm was experimentally validated. A complete segmentation of different planes in the environment was also successfully accomplished.