Road Detection And Tracking Techniques For Visual Navigation

With the development of associated hardware and software technologies, intelligent vehicles based on intelligent driving assistance and intelligent security technology have gained great attention in most countries around the world. Especially, visual navigation is one of the core technologies of the intelligent vehicle navigation. As an important part of visual navigation, road detection and road tracking attracted scholars in different countries quite earlier. The research for structured road detection technology is mature at present. This article, specifically classified into the following aspects, is primarily intended for unstructured road detection and tracking.(1) Through analysing road scene and actual road data collected,this paper pre-processes road images according to the complexity of unstructured roads.(2) Based on the comparison of road features in the different color spaces, this article designs two kinds of road segmentation:threshold segmentation method based on the HSI color model and region growing segmentation method based on the RGB model. The former method overcomes the interference of shadows, while the latter has better results in wild environment or on wet roads.(3) Comparing advantages and disadvantages of traditional edge detection operators,we select the adaptive canny operator which is accurate in positioning, and highly effective in anti-jamming. A fusion road boundary detection algorithm is designed for segmentation and edge featuring, which uses regional road information to eliminate the false road edge calculated by edge detection.and edge information to emphasize the road boundary generated by segmentation, and then achieves the road edge detection.(4) According to the characteristic that the road boundaries are approximately straight lines in a short distance ahead of the vehicles, this papter selects the Probabilistic Hough Transform to achieve straight line fitting, integrates Kalman filter to estimate the parameters of the line, and achieves the track of the road lines.VC++ and Opencv are used to realize the algorithm. The algorithm is proved to be effective through experiments made on real road videos. The results of the simulations indicate that this system has some improvements in detecting complicated roads which are affected by shadows and spots of water, and therefore, successfully detected and traced unstructured roads.