Cascaded cast shadow detection method in surveillance scenes

摘要

Moving cast shadow detection is an active research field with improving the performance of moving object detection. In this paper，we propose a novel cast shadow detection method by exploiting the local texture similarity and the intensity ratio effectively between the shadow and background patches. Firstly， local texture analysis of the initial moving patches is carried out using local binary patterns， and a part of moving object pixels and shadow pixels will be detected exactly. Then， the residual moving pixels are classified by the intensity ratio. Consequently，moving shadow pixels and moving object pixels are located respectively. Specially，post processing is employed to adjust error pixels acquired through the first two steps. In order to prove the efficiency and feasibility， comparable experiments are performed and the results validate that it can detect the shadow pixels more accurately and completely than some state-of-the-art methods.* Corresponding author at: School of Mathematics and Statistics，Northeast Normal University，Changchun，China.** Corresponding author.