Tide Beach Modeling Based on Surveillance Video

This paper presents an adaptive background update method for waves extraction based on the physical properties of tide. Most of the previous background subtraction methods based on analysis of regional connectivity detect a wave period using a static background templates or a feature set of moving objects. However, random wave in such tide videos are not extracted well. In this paper, we assume that the effect of target extraction is due to the quality of the background that is obtained from the background subtraction. Thus, a background update method is applied with update of weighted cumulative image and a timing optimization. A novel thought based on the integral on the frame and a dynamic property of the wave is used. The results of experiments on videos taken from different angles show the validity of our method.