Detecting Algorithm for Moving Objects Based on Bayesian Judging Criterion

This paper considers the problem of accuracy for judging threshold under the complicated circumstance. In the detecting system, threshold is one of the most important factor, it decides the accuracy of the detecting result. Because the circumstance is changing, the threshold is asked to adapt the change. The traditional algorithm can hardly satisfy the need of the system. Bayesian model is an efficient system based on statistics rule, and it can give a better detecting result. In order to adapt the change of the light in a same video sequence, Bayesian judging criterion is used to detect object, void warm price and falling report price is considered comprehensively, combined with likelihood function and Bayesian risk assessment, an adaptive threshold is obtained. The threshold is determined by mean and variance of the image, so it is an optimal threshold changed with every image. The optimal threshold is used to separate object from background. Compared with the traditional threshold, it can suit different circumstance. The experimental result shows that the background noise can be removed with the dynamic threshold and the moving object can be detected accurately.