Research On Road Detection Technique In Vision Assistant Navigation Of Intelligent Vehicle

Road detection technique is a very important part of vision navigation, which is the key technique of intelligent vehicle guidance. This paper analyses and researches the methods of road detection based image process. In this paper, research background, meaning and current status of vision navigation are firstly introduced, and the analyses of current road detection algorithms are presented. Since single algorithm is hard to have best behavior in all kinds of scenes due to the influence of the variational surroundings, the road detection method based on the classifying of road scenes and reconfigurable computing technique is studied. After dividing the road scenes into three classes by the characteristics of road images, the detection algorithms for each scene are designed, and the experiment system chooses the best algorithm for current scene by evaluation condition .In the paper, reconfigurable computing technique is adopted to realize this method in real-time, and Handel-C advanced language and DK compiler are used to design and exploit the reconfigurable device. The PC control module controls the whole experiment system, which is developed by VC. By applying to process 173 actual road images under several scenes, the experiment validates the method proposed by the paper, and compares it with the single self-adaptive algorithm. The method obtains right rate 86.13%.