Pixel-based heterogeneous traffic measurement considering shadow and illumination variation

This paper endeavors to develop a new background estimation model for foreground segmentation using traditional background subtraction technique. Particularly, incorporation of luminance- and pollution-controlling parameters enables the model to address illumination variation and tail backs from vehicles, respectively. To classify each foreground pixel, a new heuristic dynamic threshold-difference function is proposed for determining individual...