Research On Lane And Pedestrian Detection For Intelligent Vehicle Vision Navigation

Vision navigation is the key technique of intelligent vehicle, and the lane and pedestrian detection is the most important technique. In this paper the algorithm of lane and pedestrian detection based image process is researched. First, the development of intelligence vehicle in the word is introduced. Then, some companies and research institutions that are related to the intelligence vehicle are listed to show their products of intelligence vehicle, the development of machine vision is described, and the obstacle algorithm analyzed.Second, the monocular vision video sequences are obtained by a hand-held video camera used to replace the video camera equipped on the moving vehicle, and we use vision recognition algorithm as the main processing algorithm, PC as the main hardware platform. The system will be used to recognize the status of front road, recognize the variant information of the lane and the pedestrian and then send the processing results to auxiliary system as a main information source. The vision algorithms in this dissertation include two parts: the structure lane and the pedestrian detection.In the structure lane detection, primarily studies the problem about recognizing the edges of the road for navigation and adopts following operations sequence: Image obtained→Image preprocessing→Road edge detection→morphology dilating. For Image preprocessing, we use the method of gray level to extract the road region, in the edge detection, analyze the different algorithm and then choose sobel to detect the edge of the road, then use morphology dilating to smooth the edge. Finally, the edge is restituted , finding they have the good coherence ,and this testifies the method to be right.In the pedestrian detection, this dissertation is focused on the pedestrian detection in the monocular vision grayscale video. Firstly, the methods of the motion detection of pedestrian in video images are analyzed. Then a new algorithm to detect the pedestrian is propose, first we use the filtered background and the current frame to do subtraction, and then do image segment on the bw frame, and use morphology dilating and filling, and complete the pedestrian detection.