Obstacle Detection In Vision Navigation Of Intelligent Vehicle

In visual navigation of intelligent vehicles system, how to effectively avoid the obstacle in front of the road, to achieve the automated driving technology, the obstacle in front of the road should be detected out. It is a very important technology in visual navigation of intelligent vehicles system. In view of the vehicle running environment of structured road, the roads and the obstacle detection technology is researched, in order to realize the auxiliary driving function, effectively implement obstacle detection.The obstacle detection platform can be established with image processing and ranging principle of binocular stereo vision in the visual navigation of intelligent vehicles system. This platform can be divided into two parts, the first part is the hardware part, and the second part is the software portion. Two cameras are placed in parallel and optical detecting means is formed as part of the first, and the optimal arrangement of binocular vision system should be researched. In the software portion, it uses improved inter-class variance method to find the optimal threshold from the target image which acquired by binocular stereo vision system. De-noising processing on the binary image should be taken by morphological operations in order to obtain ideal result. Afterwards, find the road by the least-square linear fitting method, finally, extract, the whole road surface by boundary tracing. According to the ambient background on the road and the main obstacle on the road being running vehicles, symmetry characteristics of the obstacle detection method should be used in order to attain the position information of the obstacle, and to achieve the ends of detecting the obstacle. The optical imaging system of binocular stereo vision should be demarcated in order to detect the forward obstacles accurately. Furthermore, the three-dimensional information of the objectives is recovered by imaging principle and traffic information. Using the triangulation, the distance between the obstacle and the car body can be calculated. Additionally, the error of the ranging results should be analyzed.The image which acquired by binocular stereo vision system can be processed for real-time and the simulation of the correlation algorithm can be analyzed using matlab. The road sections and obstacles can be effectively divided combining the threshold segmentation with the morphology processing. The obstacle can be detected effectively with symmetry quantitative calculation method. Eventually, the purposes of the intelligent vehicle autopilot and effective obstacle avoidance can be implemented.