Motion object tracking algorithm using an improved geometric active contour model

A motion object tracking algorithm based on improved geometric active contour model is proposed in this paper. Firstly, statistical features of inter-frame difference are used to detect the bounding rectangles of moving objects, which can be set as the initialization for contour fitting. Then by introducing an enforcement to level set function, an improved geometric active contour is implemented to obtain the accurate contour of objects. Lastly, Kalman filtering is performed for motion predicting and tracking. Experimental results show this method is effective and accurate.