Identifying moving objects in a video using modified background subtraction and optical flow method

摘要

To identify the object in a video that changes its position from time to time in a video is amongst the crucial steps in various application of computer vision like extraction of objects, video surveillance, classification and pattern identification as well. In this paper a novel approach is proposed that is used to detect the objects that are not stationary in noisy environment. In the proposed algorithm Gaussian mixture model is used to extract the moving object and then the Optical flow method is used to acquire error free results quickly followed by frame differencing to obtain accurate extraction of shapes. The algorithm has been analyzed and tested over various videos in different environmental circumstances. The result acquired proves the algorithm to be more reliable, the detection is more precise and even low in terms of cost and memory requirement as well.