基于单目视觉的微型飞行器移动目标定位方法

针对目标在地形高度未知环境中移动的情况，给出一种利用微型飞行器机载单目摄像机进行目标定位的方法。首先,借助光流直方图从当前图像帧中提取出移动目标局部区域内的背景特征点|然后,结合机载微机电系统(micro electro mechanical system, MEMS)/ 全球定位系统(global positioning system, GPS)传感器测量的飞行器位姿和空间平面点成像的单应变换关系，在期望值最大化算法中将背景特征点分类为辅助平面点和非辅助平面点，并估计辅助平面到摄像机光心的距离参数和法矢量参数，从而确定移动目标所处辅助平面的空间平面方程|最后,联立求解目标视线方程和辅助平面方程获得交点坐标，转换到惯性系下完成移动目标的地理定位。实验结果表明，当微型飞行器飞行高度为100 m时，操作人员单次点击移动目标的定位误差在15 m以内。

Moving target geolocation for micro air vehicles based on monocular vision

Aiming at the movement of the targets in unknown altitude terrain,a monocular camera based target geolocation method for micro air vehicles(MAV) is presented.Firstly,the optical flow histgram algorithm extracts background features in the target’s local region.Secondly,these features are clustered into two possible classes including aided plane features and non-aided plane features by the expectation maximization algorithm,in which the homography relationship between MAV’s flight status measured by onboard micro electro mechanical systems(MEMS)/ global positioning system(GPS) sensors and planar is used.Meanwile,the normal vector of aided plane and the distance between the camera and the plane are estimated.Then the aided plane equation can be establised.Finally,the moving taregt can be geolocated by calculating the intersection of target’s sight line and aided plane in inertial frame.Experimental results show that this method can instantaneously geolocate the moving target by operator’s single click and the error can reach less than 15 m for an MAV flying at 100 m above the ground.