Evaluation of motion detection techniques for video surveillance

Video motion detection is fundamental in many autonomous video surveillance strategies. However, in outdoor scenes where inconsistent lighting and unimportant, but distracting, background movement is present, it is a challenging problem. Recent research has produced several background modelling techniques, based on image differencing, that exhibit real-time performance and high accuracy for certain classes of scene. The aim of this paper is to assess the performance of some of these background modelling techniques, namely the Gaussian mixture model and the hybrid detection algorithm, using video sequences of outdoor scenes where the weather introduces unpredictable variations in both lighting and background movement. The results are analysed and reported, with the aim of identifying suitable directions for enhancing the robustness of motion detection techniques for outdoor video surveillance systems.