A fast image rectification method for video captured with UAV hovering over traffic intersections

With the development of UAV (Unmanned Aerial Vehicle) technologies, video clips recording high-resolution ortho-images covering intersections can be captured with video cameras mounted on a UAV at a given time and location. From these clips, traffic flow information can be derived. However, since UAV is easy to be affected by external atmospheric conditions, images in video clips may be jitter or unstable, which poses a challenge to automatic information extraction. This paper aimed to present a fast image rectification method to rectify geometric information for continuous images in UAV video clips. The proposed method is a feature-based approach especially for video clips covering traffic intersections with multiple groups of zebra lines. It consists of the following three steps: (1) zebra lines detection, (2) feature points calculation and (3) perspective translation. This approach is implemented and applied to an intersection video clip, and results demonstrate that the proposed method is able to correctly rectify most images in the clip and the rectifying rate is about 12.5 images per second.