부산대학교 도서관

As for heavy traffic, road congestion and the hard extraction of vehicles in dynamic video, an improved method of dynamic video background updating is proposed. By using frame-to-frame differences and mask technology, the hole of difference image is eliminated, then extracts the reliably initial background. The image is segmented by using double threshold based on the combination of Bacterial Foraging Optimization Algorithm and Otsu Algorithm to improve the effect and reliability of background updating. Experiments shows that it is an effective way that is good in speed and effective in background extraction, and provides a good foundation for further study of the vehicle tracking.