基于主动视觉的智能车导航系统研究-控制理论与控制工程专业论文.docxVIP

PAGE PAGE 2 Abstract The intelligent vehicle is a special transportation system which is more safe and makes less traffic jams. It is becoming more and more significant in urban traffic system. Among all the navigation modes, vision navigation is worldwide researched because of it’s similarity to the human-driving. Vision navigation performs very well on highway, because the highway is structuring and has small-curvature roads. The environment of the highway is also relative simple. However, the vision navigation is facing a lot of challenges when it is used on urban environment. Not only does the urban environment have different kinds of road, but also has a lot of sharp turns. There is always no road in the image captured by the camera when the vehicle is turning. An active-vision based intelligent vehicle navigation is proposed in this paper, which aims at solving the problem of vehicles in urban environment. The active PanTilt unit is designed, which could turn the angle of the camera just like the drivers turns their head and eyes when driving. This could widen the visual field of the camera without the interruption image and image distortion. It is proved to be more reliable and accurate in road tracking. The active PanTilt unit is the foundation of the active vision navigation.A new kind of Active PanTilt control system based on Freescale S12 singlechip is designed, which uses CANopen protocol as the communication protocol. A multi-membership function based fuzzy control is used to avoid the oscillation of the camera angle. The road detection is the key factor of the vision navigation. A random Hough transform method is used. First, the image captured by the camera is operated by the Sobel Operator and becomes the edge image. Second, straight lines in the edge image is detected by the random Hough transform. Several constraint conditions is useful to find the road among the lines. According to the position of the road in the camera, the PanTilt unit is tur