Autonomous 3D modeling for robot arm based scanning

Abstract

Precise 3D modeling of a workpiece is essential in many industrial applications, such as workpiece localization and inspection. However, the limits of sensor fields of view (FoV), object self-occlusions, and robot arm footprints demand a proper selection of scanning paths and views. In this paper, we propose an autonomous modeling scheme using a next best view (NBV) algorithm based on an adaptive structured light system, where projection patterns are adjusted in real time to achieve high-resolution point clouds with high efficiency. Moreover, a probabilistic method is developed to utilize the surrounding environment information for collision-free path planning. Our results show the system can achieve a 3D reconstruction of workpieces with better completeness, quality and a less working distance in total.