Road Detection And Tracking Based On Monocular Vision

Recently, coupled with the development of the world's economy, theexisting transportation system can't meet the needs. Many countries andinstitutions have invested so much into the intelligent transportation system(ITS).The most important technology of the ITS is the vehicle assistant driving andautomatic guidance.Because the vision guidance system can get wide and integrate vision, costsfewer, and can fit the perceiving way of the human, this paper applies monocularvision method into the vehicle assistant driving and the automatic guidance. Atpresent, although there are many vision guidance systems, none of them canmeet our requirement in real-time, robustness, and practice.Firstly, this paper analyzes the structure of the vehicle assistant driving andthe automatic guidance system, and the emphasis is road detection and tracking.Compared with reappearing three-dimensional scene by stereo vision, itdemonstrates the merit and practical significance of the monocular visionmethod.Secondly, based on having analyses the character of the road, it builds thegeometry model of the road and the vehicle model when driving and themeasurement model of the camera's collecting images, and the road and vehiclemodel's stability is simulated ,which is ready for using Kalman filter to track theroadway.At last, based on the current clustering algorithms, it adopts an improvedhierarchical clustering method to track the roadway, and uses Hough transformto extract the present lanes according to the character of the road. It is indicatedthat this system satisfies the requirement of real-time and has good robustness.After that, this paper describes the principle of Kalman filter and how to useKalman filter to track the roadway.An easy method of detecting and tracking roadway is built ,which can be agood reference for vision-based guidance research.