An Improvement on Cast Shadow Detection for Vehicles Through Separately Locating Sub Shadow Regions

This paper makes an effort on locating sub-shadow regions by the inherent feature of each in order to increase the detection accuracy. We will present a study on dividing the cast shadow region into three distinct sub regions, and then separately locate them through their gray level and color features. For gray level detection, we calculate the distribution of gray level reduction rate compared to background. For color information, we discuss the disadvantage of classic methods to threshold the color components separately, and propose a new detecting algorithm using the color shift angle. Finally, we point out a pitfall situation for classic methods, and solve the problem by combining the two examinations on each sub region