Computer Vision Analysis for Vehicular Safety Applications

In this paper, we present our research on using computer-vision analysis for vehicular safety applications. Our research has potential applications for both autonomous vehicles and connected vehicles. In particular, for connected vehicles, we propose three image analysis algorithms that enhance the quality of a vehicle’s on-board video before inter-vehicular information exchange takes place. For autonomous vehicles, we are investigating a visual analysis scheme for collision avoidance during back up and an algorithm for automated 3D map building. These algorithms are relevant to the telemetering domain as they involve determining the relative pose between a vehicle and other vehicles on the road, or between a vehicle and its 3D driving environment, or between a vehicle and obstacles surrounding the vehicle.