In many vision based application identifying moving objects is important and critical task. For different computer vision application Background subtraction is fast way to detect moving object. Background subtraction separates the foreground from background. However, during background subtraction pixels belonging to shadow are misclassified as foreground object. Moving cast shadow associated with moving object also gets detected making it challenge for video surveillance. Now days many methods are available for background subtraction. The core of background subtraction is background modeling. Gaussian Mixture model is good balance between accuracy and complexity. However, it is difficult to determine the optimal color space in which to remove shadow. In this paper, we study the features of moving object and shadow in different color spaces to solve the problem.

No abstract