Object location prediction based on motion estimation and traffic density on digital video

This paper discusses a method for predicting the location of foreground objects in a video. The method utilizes both the object motion and statistical traffic density to predict the future location of the object. The object motion is estimated from the past video and the traffic density is obtained by analyzing the historical results on the video. The key component of the prediction method is the adjustment of processing area of the classifier. This adjustment mechanism greatly improves the detection efficiency in terms of computational cost in the foreground-detection task. The proposed algorithm is experimentally tested on three different scenarios. These experimental results demonstrate the advantage of using the proposed prediction method.