Robust Small Moving Target Detection based on Adaptive Structure Constraint in Infrared Sequence

The detection of small moving target under the complex ground background still remains a challenging task, because of small size of targets and large dynamic change of backgrounds. In this paper, we present a novel moving target detection method based on adaptive structure constrain t(ASC) in infrared sequence. Firstly, the DOG filter is used to extract the singular points of the first frame, which are considered as a candidate points set. Secondly, the kernel correlation filter-based method is adopted to track the points set on consecutive frames. The corresponding points set between the first frame and the last frame is established. Thirdly, an adaptive structure constraint is proposed based on the corresponding points set to suppress the background movement. Finally, we explore a threshold segmentation method to extract real moving targets. The extensive experiments demonstrate that the ASC outperforms other baselines in complex air-ground scenes.