Automatic detection technique of preceding lane and vehicle

Abstract

An automatic detection method of preceding lane and vehicle for structural road is proposed. Lane model is dynamic switched for different road, and then the rapidity of linear lane detection and the accuracy of curve lane detection are achieved using Kalman tacking technique. The initial vehicle detection region is dynamic determined with shadow detected by adaptive double-threshold method. Then vehicle hypothesis is generated by edges detected with improved SUSAN algorithm, as can enhance the speediness of vehicle detection. The experiment results show that the algorithm is feasible for different lane models, and can detect preceding lane and vehicle effectively.