3D Vision for Large-Scale Outdoor Environments

Abstract

The difficulty of inferring depth from a 2D image has resulted in a growing interest in rangefinding technologies. This paper describes the development of a low-cost 3D vision system that was used to extract a rich data set from an unstructured outdoor domain. The system consists of a laser rangefinder (ladar) and monocular camera, both mounted on a pan-tilt unit for accurate positioning. By sensor fusion of ladar data with still images captured by the camera, a high-resolution environmental model was created that represents surface shape and colour. Preliminary results in proximity, surface normal directionality, and colour based scene segmentation through semantic-free clustering processes, are presented. In the context of robotics, a future project will utilise this new sensor system for the purpose of natural landmark based localisation in rough terrain.