Research Of Road Identification For Outdoor Vision-Based Autonomous Navigation Vehicle

Vision-based navigation technology is a new hot spot in the area of autonomous navigation vehicle research. Images describe roundly the environment, and carry a tremendous amount of information about the scene. It is the important artifice for vehicle to explore the surroundings. The article mainly aim to the application of vision system in outdoor non-structurization environment, and emphasize discussed the camera calibration,road edge detection and the image segment based on edge and color.Before the calibration, the article first research the installation of the camera, and a scheme of the camera installation under heeling angle is purposed. On the basic of consider the camera internal and external parameters and radial distortion, utilization the ground grid as the calibration template, and using the Tsai two-step method to calibrate the internal and external parameters, and finally use the experimental results validate the above conclusions. The result indicated that the installation scheme and the algorithm is feasible.Because of the image have noise and must to carry through the image preprocessing, so the gray level transformation,histogram handling and airspace filtering are used in this article. Then get across comparison several edge detection operators, choose the Sobel operator to detect the road edge. In order to detect the road-edge straight line, the article is on the basic of analysis the Hough transform algorithm for line detection and the organizing into groups algorithm for line detection, analyzing the advantage and disadvantage of each algorithm, combine the advantages of them, purposed the use of generalized Hough transform based on organizing into groups in road edge straight line detection. The experiment indicated that algorithm for line detection has high computation speed and strong robustness.When processing the image segmentation, select the auto-threshold method, adopt the maximum area method to choose the initial threshold, application the max ratio of class among variance and class internal variance to carry through the initial segmentation. To settle the problem of shadow's disturbing in road identification, three special color spaces are made. First to carry through the initial segmentation in brightness space, then according the characteristic of the shadow to detect the shadow in the other spaces, and use the edge to carry through image segmentation, and the last to process the final image fusion and choose the approve running area of the vehicle.