Extraction and tracking of moving objects for smart surveillance system

The problem of extracting and tracking moving objects in static background scene by color video is addressed. The algorithm utilizes normalized rgb space combined with gray space model to immune the effect of shadow upon detection. A GMM model of difference image is constructed. Based on the model, a motion detection operator is introduced to generate a motion border image. GVF-Snake is enhanced to extract the contours of moving objects in video sequence by modifying the energy entry and adding the function of automatical contour initialization and position prediction. This algorithm is proved to be effective for both rigid and non-rigid objects.