An Adaptive 3D Panoramic Vision System for Intelligent Vehicle

Advanced driver assistance systems (ADAS) improves the driving safety of vehicles greatly, and the vehicle-mounted panoramic vision system is one of the most important type of ADAS. Because of small observation range and fixed area of interest, the existing panoramic vision system can't provide satisfied performance. In order to solve the defects above, an adaptive 3D panoramic vision system for intelligent vehicle is proposed. By the method, multiple sets of cameras are installed around the vehicle to get driving conditions around the vehicle firstly. Then the image distortion is amended by image preprocessing, and the front and back frames of same camera are fusioned and spliced to further eliminate the visual blind area of the vehicle. By the approach of target detecting and anti-collision warning, different levels of dangerous targets in the driving process are given corresponding warning prompts. Finally, to the three-dimensional panoramic vision is generated by the three-dimensional reconstruction and texture mapping. In order to satisfy the drivers' different areas of interest to the vehicle environment at different driving speeds, this paper proposes three-dimensional model adaptive and observation angle adaptive transformation methods to provide different three-dimensional models and observation angles at different driving speeds. The simulation results show that the proposed system solves the problem of large image distortion and small observable area in the 2D panoramic vision, and the problem of poor adaptability of the existing 3D panoramic vision system model and observation angle is updated also. Then the driver's sense of observation can be improved and the vehicle driving safety can be enhanced correspondingly.