Application of Background Estimation on Integrated Inter-frame Subtraction and Clustering Statistic in Tunnel Environment

摘要

The current background estimation methods inadequate with many disadvantages. Firstly it cannot satisfy the requirements of the background reconstruction and updating under specific environment. Secondly it involves the problems about both computation inefficiency and space insufficiency. This work proposed a background estimation method on the basis of integrated frame subtraction and clustering statistic in vehicular tunnel environment. The method is to carry out cross-subtraction for video sequence within a certain duration range by interframe subtraction to eliminate redundant noises, and choose the most appropriate pixel point in the video sequence to finalize the background reconstruction effectively, and then complete the background updating through the analysis of changing situation near the peak point of subtraction image histogram. The experiment result demonstrates proposed method can achieve background reconstruction of vehicular tunnel under different traffic conditions with only a little image frames, which can significantly reduce not only system computation complexity but also memory usage. In addition, background updating can be robust completed under the situation of camera shake and lighting changes, meanwhile the proposed method can easily become a real-time practical application.