无人直升机自主着舰的光学3D测量解析算法

推导了一种用于无人直升机自主着舰的相对位姿单目单帧视觉测量算法。设计了基于正方形主特征的着降区域平面特征图案,根据几何透视投影学理论,采用摄像机针孔模型,以特征图案主正方形四角点的物理坐标和图像坐标作为输入,并以正方形邻边的垂直特性作为束缚条件,推导出了位置和姿态6个参数的解析表达式。通过在实验室环境下对真实图像的测量表明,当特征图案主正方形边长为0.1m(等效实际甲板特征图案边长1m)和相对距离为1m(等效实际距离10m)左右时,位置参数RMS误差小于0.2cm(等效为2cm),姿态参数RMS误差小于1.2°,算法具有很好的稳定性,可用于无人直升机着舰过程中的位姿测量。

Vision based analytic 3D measurement algorithm for the autonomous landing of unmanned helicopter on ship deck

In order to improve the performance of relative 3D measurement in the autonomous landing of unmanned helicopter on ship deck,a novel algorithm based on monocular vision and single image frame is proposed for the high level parameters estimation in the image processing.A landing mark including one main square is designed to contain the information needed for the 3D parameters calculation.The theory of perspective projection and the pinhole model of camera,as well as the constraints of vertical relationship between the adjacent sides are used in the algorithm's deducing.With the world coordinates of the square's 4 corner points and their corresponding image coordinates as inputs,the analytic functions of the position and orientation parameters are obtained.The tests on real image show that the state estimations are accurate to within 2cm in each axis of translation and 1.2°in each axis of rotation when the equivalent distance between the helicopter and the ship deck is about 10m,and the algorithm is effective for the control of unmanned helicopter automatic landing on ship deck.