A Fast Vision-Based Air Object Detection Algorithm

Abstract

Most of air object detection techniques are based on laser or radar sensor. This paper addresses a novel technique for air object detection which is based on computer vision, namely it detects object with sequence images acquired by visible or thermal infrared cameras. It is based on single frame detection other than classical background subtraction, which means that it can tolerate motions or jitters of a camera and sudden changes of illumination. Moreover, it can be used both in visible-light and infrared images. Extensive experiments demonstrate the effectiveness and efficiency of the proposed algorithm. It has been successfully used in movable surveillance system.