Moving object detection using motion enhancement and color distribution comparison

To detect moving objects from videos taken by unmanned aerial vehicles, an approach using motion and color cues was proposed. It adopted the forward motion history image to enhance the independent moving information and suppress the noises in background areas, enabling the candidate moving regions to be extracted entirely. Then, an iterative local color distribution comparison based the method was developed to remove the background pixels contained by each candidate region, insuring each moving object to be extracted precisely. Thus, the cost of computation is saved, and the miss rate and false alarm rate can be cut down prominently. Experimental results with several aerial videos demonstrate the performance and the robustness of the proposed method.