Cast Shadow Detection for Surveillance System Based on Tricolor Attenuation Model

Shadows bring some undesirable problems in computer vision, such as object detecting in outdoor scenes. In this paper, we propose a novel method for cast shadow detecting for moving target in surveillance system. This measure is based on tricolor attenuation model, which describes the relationship of three color channel’s attenuation in image when shadow happens. According to this relationship, the cast shadow is removed from the detected moving area, only the target area is left. Some experiments were done, and their results validate the performance of our method.