A Four-Frames Differencing Technique for Moving Objects Detection in Wide Area Surveillance

Abstract

Moving objects detection or change detection in video sequences, is a fundamental task in video surveillance applications. Although, existing methods perform well on videos filmed by stationary cameras, these methods fail dramatically in videos filmed by non-stationary cameras. In particular, in very low frame rate and sudden illumination change scenarios, like Wide-Area Motion Imagery (WAMI). In this paper, we show the limitation of the widely used three-frames differencing technique in such scenarios. Then, we propose a pixel-level algorithm based on four-frames differencing rather than three-frames differencing, which contains the temporal information needed to separate the foreground (moving objects) from the background robustly. We compare our method to the state-of-the-art methods in detection quality. Experimental results on the WPAFB2009 dataset show that our method outperforms existing methods.