Contributions to intelligent visual surveillance

点击翻译按钮获取中文摘要

A recent analysis of the needs and challenges of intelligent video surveillance has shown that current commercial systems lack many useful video processing features because the works developed in research are still not producing results that are reliable enough. Thus, actual commercial systems only rely on basic image processing, do not assist well security agents, and as such cannot be considered as being intelligent. Among the video processing features needed for intelligent video surveillance, we can find 1) human tracking, 2) tracking with Pan Tilt Zoom (PTZ) cameras, 3) video fusion from different types of cameras, 4) the use of thermal sensors, 5) crowd analysis, and 6) understanding the activities in the scene. This research program aims at studying the first four required video processing features.