Motion detection for video surveillance

Motion detection is one of the key techniques for automatic video analysis to extract crucial information from scenes in video surveillance systems. This paper presents a new algorithm for MOtion DEtection (MODE) which is independent of illumination variations, bootstrapping, dynamic variations and noise problems. MODE is pixel based non-parametric method which requires only one frame to construct the model. The foreground/background detection starts from second frame onwards. It employs new object tracking method which detects and remove ghost objects rapidly while preserving abandon objects from decomposing into background. The algorithm is tested on public available video datasets consisting of challenging scenarios by using only one set of parameters and proved to outperform other state-of-art motion detection techniques.