Vision Based Localization and Environment Modeling for Unmanned Vehicles

Safety and operational demands require that operators of unmanned security and defence vehicles be located at safe distances. The capability of creating photorealistic 3D models using on-board sensors on unmanned vehicles will improve the operatorsí situational awareness. Over the last years MDA in Brampton have been developing computer vision technologies for localization and 3D modeling of environments and objects from mobile cameras. The technology relies on real time image acquisition and processing using a combination of dedicated hardware and software. The basic system processes images from stereo cameras and computes the camera motion in 3D (position and orientation). A real-time version has been deployed on a mobile rover and has been used as a visual odometry sub-system for off-road navigation. The on-going work focuses on augmenting iSM with additional sensors (Infrared and Gamma) and creating multi-modal representations of environments contaminated with Chemical, Biological, Radiological and Nuclear agents.