An adaptive background subtraction scheme for video surveillance systems

The effectiveness of most of the traditional background subtraction schemes employed in video surveillance systems are dependent on proper selection of a threshold. In an endeavour to avoid this dependence, this paper proposes a novel motion detection scheme for video surveillance systems, based on an Eigen background subtraction scheme with adaptive threshold. The performance of the proposed algorithm has been tested using a standard data set. The results achieved using the new method has been compared with that of a conventional Eigen background subtraction method and a Gaussian mixture model scheme. Enhanced object detection accuracy has been observed with the proposed scheme. The improved detection efficiency has been attributed to the adaptive threshold introduced in the new method for determining the background pixels.