Comparison of Performance of Different Background Subtraction Methods for Detection of Heavy Vehicles

Abstract

The growing vehicle numbers in urban and national road networks emerged the need for effective monitoring and management of road traffic. Especially detecting vehicles with break average speed limits rules and trespassing a heavy vehicle is essential to constitute safety traffic flow. In the proposed study, the main goal was detecting heavy vehicles using surveillance videos by using interframe difference, approximate median filtering and Gaussian mixture models for background subtraction and compare their performance. Moreover, after removing the background image from original videos, on binary image morphological opening and blob analysis processes were applied and with minimum blob area of the detected object in a frame, heavy vehicle detection was achieved. Different background subtraction methods produce varying results, and these results were discussed. Our results were consistent with performance comparison studies which indicated the Gaussian mixture model was stable, real-time outdoor tracker in any varying outdoor condition.