基于视觉的无人作战飞机跑道障碍物检测方案

Vision-based runway obstacle detection scheme for unmanned combat aerial vehicles

提出了基于视觉的无人作战飞机跑道障碍物检测方案.结合应用特征和流的障碍物检测方式的特点,使用多尺度特征点匹配光流估计方法代替普遍使用的微分光流计算方法,直接对图像序列计算稀疏光流场;利用相关假设,在存在一定导航误差的情况下,对跑道障碍物进行实时检测;同时对跑道障碍物视觉检测方案的误差和适用区间进行了分析;通过自主开发的"无人作战飞机自主着陆实时仿真验证平台"进行仿真,结果显示该方案能够有效检测跑道上存在的障碍物.

A computer-vision-based runway obstacle detection scheme for an unmanned combat air vehicle （UCAV） was presented. The scheme combined the advantages of the feature-based and the flow-based obstacle detection algorithms. Instead of using gradient-based method, a multi-scale optical flow estimation method based on feature point matching was adopted, which make it possible to calculate sparse optical flow field directly from image sequences. Under some relative hypothesis, obstacle on the runway could be detected even with certain navigation errors. The detection sensitivity and the stage applicable for obstacle detection were also discussed. The obstacle detection scheme can run properly on the real-time simulation system for autonomous landing of the UCAV.