Motion detection with using theory of entropy

The traditional automatic smart image surveillance system can usually be used in the environment with still background. That is, the background image must not contain the moving objects. If there is waving ocean, waving tree, floating cloud, or raining in the background image, the traditional methods do not work well. In order to improve this problem, a new motion detection method based on the theory of entropy and combined a multi-periods sigma-delta background estimation algorithm is developed in this paper. Based on the theory of moving average, a moving thresholding method is designed in this paper to obtain a sequence of alarm announcements. Experiments are carried out for some samples with dynamic backgrounds to demonstrate the computational advantage of the proposed method.