Moving shadow detection based on spatial-temporal constancy
Andre Russell, Ju Jia Zou · 2013
An accurate estimation of the target object size and shape is an essential step towards many computer vision applications such as object tracking and object recognition. Due to the presence of cast shadow, these properties cannot be extracted using ordinary object detection systems. This paper introduces an effective method for detecting moving cast shadows by exploiting spatial and temporal color constancy among pixels. Using an initial clustering of the current frame, spatial and temporal color constancies are checked for each region to classify as shadow those with similar constancy patterns. The advantage of this technique is its capability of detecting cast shadow when having foreground-background camouflage. In addition, it can detect cast shadows in various environments for both indoor and outdoor sequences. Experimental results show higher performances of the proposed method over other methods and better achievement in terms of detection rate and shadow discrimination rate.