Background modeling method based on improved multi-Gaussian distribution

Abstract

For motion detection based on background difference method, there is a method of estimating background by multi-Gaussian distribution model. But in this method, the update rate alpha for both background model and the weights of Gaussian model affects directly the speed of background modeling and capability of resisting disturbance in modeling. A new method of adaptively changing update rate alpha is put forth in the different stages of the background modeling. In this method, the endpoints of background model established in the first stage are found out firstly in the study of some frames at the beginning of video, by taking the contrast ratio of real-time background images as standard of preliminary establishment of background model; then, change alpha is changed and the background model is solved, till establishment of background model is completed. The experiment result shows that compared with the traditional method, the improved algorithm proposed in the Paper has a better capability of resisting disturbance at the same time of ensuring rapidity. This method can well satisfy the requirement of real-time system, which lays the solid foundation for accurate detection of the subsequent motion target.