Research On Obstacle Detection And Identification For On-board Vision Systems

The intelligent vehicle is a vehicle with environmental perception, pathplanning, autonomous decision-making, and automatic execution. With the development of computer technology, artificial intelligence and various of sensor technologies, the techniques of intelligent vehicles become a fast developing area in the intelligent transportation field. Automatic environmental perception is one of the mark functions of intelligent vehicles, which detects and perceives the surroundings of vehicles using vehicle-mounted sensors, where the main goal is to determine the roads, obstacles and other movement targets. Therefore, this paper further discusses the key technologies of vision-based road obstacle detection and identification for intelligent vehicles.The 2D laser scanner and infrared image system set up in this paper are about to ensure that the mobile monitoring system can be used under poor visibility or at night and the ability of recognizing barrier is improved. Based on the system, the data registration of 2D laser scanner data and infrared image data are proposed first. Secondly, considering the intelligent vehicle can be moving or still, recognizing moving obstacle system in the both motionless and moving environment based on 2D laser scanner data is set up. Finally, regions of interest are found according to 2D laser scanner, feature is extracted by sparse coding and pooling and obstacles are recognized by support vector machine. The results have shown the feasibility of the algorithm used in this paper.