Wide Range and Dynamic Video Image Stabilization Applied in UAV Aerial Photographing

摘要

In the process of aerial photographing, due to the wide range of complex scene conditions such as air flow changes and the influence of light and atmosphere, which lead to the large dynamic instability phenomenon such as mechanical vibration and even violent shaking when the camera carrier moves, this paper proposes a new video stabilization algorithm for the wide range and large dynamic state. Firstly, the feature points of reference frame are extracted by GFTT algorithm, and the motion vector between frames is obtained by using sparse optical flow method to find matching feature points in the current frame of continuous video image. Secondly, on the basis of traditional Kalman filter, a hybrid filter composed of Kalman filter and Gauss discrete filter is used to smooth the motion of each frame image. Finally, a stable video sequence is obtained by motion compensation. The experimental results show that the MSE parameters of this algorithm are 21.4% lower than those of the traditional Kalman filter based image stabilization algorithm, which can meet the requirements of wide range and dynamic video image stabilization.