基于图像匹配的无人机自主返航技术  

Vision-based UAV autonomous return

作　　者：盛哲[1] 邹丹平[1] 刘佩林[1] 赵毅 

机构地区：[1]上海交通大学电子工程系,上海200240 [2]上海华测导航技术有限公司,上海200240

出　　处：《信息技术》2016年第1期13-16,共4页Information Technology

基　　金：国家863计划资助(2012AA12A209);中国卫星导航系统管理办公室(北斗办)和上海市科学技术委员会联合资助项目(BDZX005)

摘　　要：文中基于无人机机载的视觉信息,在GPS信号不可靠的环境下实现自主返航的功能。利用无人机底部垂直向下的摄像头,对图像计算光流可以较精确地获得无人机的速度信息。在去程中,无人机每隔一定距离记录一次位移信息与向前摄像头图像的关键帧;在回程时,无人机先根据位移信息到达关键帧图像附近,然后比较当前图像与目标关键帧纠正误差,对下一帧重复这两歩直到回到起点。仿真实验证明文中所述方法能较好地实现定点悬停和自主返航的功能。In this paper, an autonomous return approach of an unmanned aerial vehicle （UAV） is proposed based on the visual information on the UAV in a GPS-denied environment. The aerial velocity of the UAV can be estimated by calculating the optical flow of the image captured by the downward-looking camera. On the go path, key frames captured by the forward-looking are record when the UAV moves a constant distance. On the return path, after the UAV reach the next frame by the estimated velocity,the relative pose between the current and the key image is estimated to help the UAV reach the recorded position accurately. Finally it returns to the start position. The experimental simulation result shows that this approach can help the quadrotor hover at one point and solve the autonomous return problem.

关 键 词：旋翼无人机 计算机视觉 自主返航 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]