The Road Recognition Based On Machine Vision Of Autonomous Unmanned Vehicle

Visual navigation is a key technology in the field of intelligent vehicle. Lane recognition is an important part of the realization of autonomous navigation while accurate real-time road recognition algorithm has become the focus of research of scholars. In this paper, In order to improve the problem of recognition algorithm that it real-time and stability is bad and low recognition rate which ours has studied the structured road detection and tracking. By theory and Experimental results show that a high real-time and good stability road detection algorithm has been proposed, which can offer accurate control strategy for autonomous unmanned vehicle. Therefore, road recognition technology research is the important significance for unmanned autonomous vehicle navigation systems.In this paper, In order to adapt to different road conditions, the paper has made the following main aspects of research for structural road:the road image is preprocessing from the camera, identification and tracking detection of road lanes, research on camera calibration technology. In order to follow the road lane real-time and accurate identification, during the road image preprocessing stage which we have done the graying of image, bilateral filtering, edge detection, OTSU image segmentation. However we filter noise and interference factor of the image, and keeping the edge information of the target image. Then identify and track the lanes with the preprocessed images. The recognition algorithm of the paper first analyzes lines of the road images, and improved Hough transform extract lane boundary points. Determine ROI (region of interest) of the lane line, and it will match lane line which adopting combination of hyperbolic model. By the least squares method to complete lane boundary reconstruction, and complete identification of the lane line. The algorithm has a better effect in straight road, but to achieve the identification and tracking of the curved road, utilized extended Kalman filter predictive tracking algorithm. The current images are used to estimate the lane position of a later time, and can identify and judge the road toward (left or right), Extracting effective road information finish lane tracking. Final, the plane calibration method is used for calibrating the camera which it is brief and accuracy. Camera parameters and parameters of the algorithm determine location of the lane line, and complete the visual understanding of road environment.Through a large number of simulation experiments to verify the feasibility of the algorithm, experimental results show that the instantaneity and stability of image processing can meet the requirements of road recognition. Different road environment recognition experiments show that recognition algorithm could quickly and accurately identify the road lane under such circumstances as a straight road, curve roads, vehicle occlusion, rainy weather, broken road lane and so on.The results of the research topic has important significance both theory and engineering application.