为了充分利用连续视觉图像中3维空间信息, 解决无人机自主降落过程中的定位问题, 在稠密3维点云法和光流法定位原理的基础上, 提出了基于同物不同时图像像空间的定位方法。以理论推算、图形注释等方式, 通过求解单个像素点和整个图像移动变化情况, 将连续帧图像的形变、量变信息分解为无人机和参照物的空间相对运动信息, 并结合已知的参照物运动参量, 推算了无人机飞行位姿信息, 完成了无人机基于光学视觉图像的空间定位方法研究。结果表明, 该研究为视觉系统在无人机降落回收过程中独立实现空间定位提供了一定的借鉴和参考。

In order to make full use of 3-D spatial information in continuous visual images and solve the positioning problem of unmanned aerial vehicles(UAV) during autonomous landing, based on the principle of dense 3-D point cloud method and optical flow method, a localization method based on the image space of the same object at different time was proposed. By means of theoretical calculation and graphic annotation, and calculating the movement of a single pixel and the whole image, the deformations and quantities of continuous frame images were decomposed into spatial relative motion information of UAV and reference objects. Combining with the known motion parameters of the reference object, the position and attitude information of UAV was calculated. The spatial positioning method of UAV based on optical vision image was completed. This study provides a reference for the vision system to independently realize spatial positioning in the process of UAV landing and recovery.