Stabilization of Video from Miniature Air Vehicles for Target Localization

摘要

Miniature air vehicles equipped with lightweight, inexpensive cameras have recently been employed in a greater variety of missions owing to their increased capability and relatively low cost. Unfortunately, miniature air vehicle video is often difficult to watch owing to high-frequency jitter caused by environmental disturbances such as wind gusts. This paper describes a methodology for producing real-time stabilized video by eliminating jitter. Video stabilization is enabled by estimating the intended video motion through the methodical selection and tracking of image features from one frame to the next. The effectiveness of the video stabilization routine is demonstrated in a real-world surveillance task of localizing stationary and moving targets on a mobile ground station. The paper also describes the unique mobile platform that is used to create and refine these miniature-air-vehicle-specific computer vision technologies.