A moving target detection algorithm based on BING objectness and background estimation

This paper deals with an intelligent image processing method for the video surveillance systems. We propose a fast and robust algorithm for moving target detection which combines Binarized Normed Gradients (BING) Objectness and background estimation. A real-time surveillance system needs to detect moving objects robustly against noises and environment. So the proposed method uses a simple background estimation model with a sensitivity parameter to extract a set of rough moving foreground in image. We also use objectness detection within the foreground set to estimate another set of candidate object windows. Then the target (pedestrians/vehicles) region by the intersection of areas derives from the former two steps. To track moving objects fast, the proposed method predicts the velocity and the direction of the groups formed by moving objects. Besides, time cost is reduced by the decrease of estimation regions. Experiments on the outdoor datasets show that the combined method can not only achieve a high detection ratio (DR) but also decrease false alarm ratio (FAR), as well as time cost. Finally, the experiments show that the proposed method has the robustness against the environmental influences and the speed, which are suitable for the real-time surveillance system.