Scene Division-based Spatio-temporal Updating Mixture Gaussian Model for Moving Target Detection

Since the traditional mixture gaussian model nonfully utilize the background distribution information in time and space, in this paper, the scene division method is used to segment the scene into the background stable regions and background disturbance regions, and a spatio-temporal stochastic updating method is proposed. Under the premise that the background disturbance areas are correctly identified as the background, the spatio-temporal stochastic updating mechanism can make the pixels in the scene have a reasonably renewal time, and then improve the detection precision of the moving target. The experiment shows that compared with the classical mixture gaussian model, the improved mixture gaussian model has the better performance of moving target detection.