A novel method for moving vehicle tracking based on horizontal edge identification and local autocorrelation images

Abstract

A novel method for moving vehicle tracking was proposed to improve the vehicle identification rate on the basis of local autocorrelation (LAC) and horizontal edge (HE) identification. Local autocorrelation images were generated as the pre-treatment for horizontal edge identification, so that the horizontal edge characteristics could be strengthened while the influence of weather conditions could be reduced. Robust background model could be obtained based on exponential forgetting method (EFM), the moving vehicle regions were detected by background subtraction. Stable horizontal edge of vehicle was detected for vehicle tracking, the length of horizontal edge was normalized in image sequence to improve vehicle detection rate. The distance of the barycentric coordinate of the horizontal edges was used to track vehicles in traffic videos. Barycentric coordinate was modified using correction coefficient to ensure the effect of tracking. The vehicle regions were marked using bounding box during vehicle tracking. Traffic videos of various complex conditions (foggy weather, strong sunlight, morning, and evening) were used as test images to verify the effectiveness of the proposed method. Experimental results show that a higher identification rate of moving vehicles is obtained via the proposed method. The proposed novel method can be used to improve the performance of the intelligent trans portation systems.