Abstract: In order to evaluate the stability of hovering UAV, this paper designs a system based on binocular stereo vision. Firstly, using the proposed neighborhood gray improved hierarchical matching strategy, it obtains the matching feature points incalibration of video images using camera calibration. Then according to the principle of reconstruction, it calculates 3D coordinate data in the view of the UAV. Finally, the stability degree of the UAV is evaluated by hovering precision measurement function. The experimental results show that the system can effectively match the feature points of the target, calculate the three-dimensional coordinates of the target, and evaluate state of the UAV hovering according to the relative control precision of the floating.

Key words: binocular stereo vision, hierarchical region matching, 3D reconstruction, hovering accuracy

摘要： 为评测悬停状态无人机的稳定性，设计了一套基于双目立体视觉的无人机悬停状态评测系统。首先结合双目相机标定对视频图像校正，并利用提出的邻域灰度改进分级匹配策略，获取匹配特征点；然后根据重建原理计算求得视野中无人机三维坐标数据；最后利用悬停精度测量函数评测无人机的稳定程度。实验证明系统有效匹配目标的特征点，计算出目标三维坐标，同时根据其相对控制精度的浮动，对无人机悬停状态进行评测。

关键词: 双目立体视觉, 分级区域匹配, 三维信息获取, 悬停精度