Vehicle detection and counting for complex weather conditions

There are many techniques used in moving objects detection. Post processing steps including setting an appropriate threshold to differentiate between the foreground and background have a big effect in increasing the detection rate accuracy of these techniques. However, up till now finding an appropriate threshold that is able to adapt itself particularly in scenes with poor visibility has not been very successful. In this paper, a new adaptive threshold algorithm based on the triangle threshold method is proposed. Together with the background modeling based on the approximate median filter, the triangle threshold is applied over the difference histogram between the current frame and the background model. Finally, morphological operations and counting are applied. We show that the triangle threshold can efficiently differentiate between foreground and background in urban roads under different weather conditions (i.e. fog and snowfall). Comparisons between the proposed method and adaptive local threshold (ALT) show the potential of our approach. The experimental results show that the proposed method has better detection rate compared to ALT while maintaining similar processing time.