A Feature Pair Based Method for Online Moving Object Detection from High-Resolution Airborne Videos

Current researches on moving object detection from airborne videos mainly focus on low-resolution videos. Recently, to achieve a wide-area of fine-grained surveillance, most of unmanned aerial vehicles (UAV) prefer capturing high-resolution videos. Traditional methods usually require the processing of many neighboring frames to achieve a good performance for detecting moving objects in a frame. However, for high-resolution videos, due to limited computing resources, it is difficult to process many frames online, making traditional methods unsuitable for this task. In this paper, we propose a feature pair based method for moving object detection from high-resolution videos. It only needs two frames to detect moving objects and most of operations are performed in the feature pair domain with high efficiency, being well suited for high-resolution situation. Moreover, we design four strategies, i.e., feature pair extraction, matched pair refinement, moving pair determination, and clustering strategies, to enable an accurate detection of moving objects. Experimental results demonstrate an excellent performance of the proposed method for moving object detection from high-resolution airborne videos.