An Approach for Enhancing the Results of Detecting Foreground Objects and Their Moving Shadows in Surveillance Video

Automated surveillance system is becoming increasingly important especially in the fields of computer vision and video processing. This paper describes a novel approach for improving the results of detecting foreground objects and their shadows in indoor image sequences. Several previous techniques have been developed in the literature that deal with moving shadows. However, a comparative evaluation of the existing approaches shows that most of the methods are unable to extract and preserve the shape of the moving objects completely. Since accurately detecting the moving objects from the background scene is a crucial step in such system, we then design a method to improve the detection results by appropriately filling the regions that are erroneously removed from foreground regions. We propose to combine the classification result of the moving objects, object representation based on the cardboard and head detection technique for performing the improvement task.