Autopilot Car Visual Navigation Theoretical Research

Lane detection and moving object detection that we studied are the key technologies in the vision navigation of the Autonomous Highway Vehicle(AHV). This paper includes the image pretreatment(edges. detection), lane detection and moving object detection.For the edges detection is the key step in the detecting of the lane, we present a new mathematical morphologic edge detection based on multi-scale contour structuring elements. The method keeps good detected edges and the details of image under heavy noise condition. Comparative study reveals its superiority over other morphologic methods and classic edge detectors.In the lane detecting, we propose a method which fuses the way of lane edges detection and the method of color-based segmentation. The method reduces the shadow and noise points influence on lane detection and is convenient in the applications in the vision navigation of the Autonomous Highway Vehicle. Imitated test reveals that our method gets a better result in different road conditions.In the moving object detecting, we propose a new background subtraction way. In our method, the combination of the Gaussian model and the least-squares in the background subtraction eliminates foreground influence on background rebuilding. Imitated test reveals that our method gets a better result than the classic ways.