From pixels to objects and trajectories: a generic real-time outdoor video surveillance system

In this paper we present an effective real-time video surveillance system suitable for a wide range of outdoor application scenarios. By carefully addressing unique challenges of outdoor scenes such as frequent illumination change and complex object motion etc., this system can robustly extract and track video objects without assuming any a priori knowledge. Through properly balancing the computation and temporal sampling, it can run in real time on a modern PC. Its superior performance is demonstrated in various complex outdoor situations.