Illegal Vehicle Parking Detection Based on Online Learning

In this paper, it is proposed an improved method to aim at the illumination vary problem for detecting illegal parking in the surveillance video. The method uses a frame to frame subtraction method for modeling the background, and update it, then using the background subtraction method to get the moving candidates, it can obtain the accurately moving objects after appropriate processing. At the same time, the work also used a method based on texture and color histogram. Besides, in order to detect the salient object, we propose a novel adaptive feature combination mechanism to combine the different features, in which the combining weight of each static map is learned using online learning. At last, it justifies whether the vehicles are parked illegally or not. Experiment results demonstrate that the proposed method significantly outperforms the other state-of-the-art methods, and it is more suitable for the real scenes.