提出一种将EPnP算法和SoftPOSIT算法融合的单目视觉姿态自动测量方法.首先, 利用EPnP算法计算得到立体靶标的位姿参数, 并将该位姿参数作为SoftPOSIT算法的迭代初始值带入.其次, 将SoftPOSIT算法计算位姿的的迭代过程与立体靶标结合, 实现姿态的自动测量, 并仿真验证了拓扑确定位姿的有效性.最后, 为了验证姿态测量结果的精度, 以高精度二维转台为基准, 将立体靶标安装在二维转台中, 通过控制转台转动角度得到靶标姿态的测量数据.实验结果表明, 当转台转动角度在［-20°, 20°］时, 靶标姿态角α的测量结果标准差为0.20°, 姿态角β的测量结果标准差为0.27°.

Aiming at the non-automatic problem of attitude measurement in monocular vision, an automatic measurement method for monocular visual attitude combining EPnP algorithm and SoftPOSIT algorithm is proposed. Firstly, the pose parameter of the stereo target is calculated by the EPnP algorithm, and the pose parameter is taken as the iteration initial value of the SoftPOSIT algorithm. Secondly, the iterative process of calculating the pose by SoftPOSIT algorithm is combined with the stereo target to realize the automatic measurement of the attitude, and the simulation verifies the validity of the topology to determine the pose. Finally, in order to verify the accuracy of the attitude measurement results, the stereo target is mounted on the two-dimensional turntable with the high-precision two-dimensional turntable as the reference, and the measurement data of the target pose is obtained by controlling the rotation angle of the turntable. The experimental results show that when the turntable rotates the angle from -20° to 20°, the standard deviation of the measurement results of the target attitude angle is 0.20°, and the standard deviation of the measurement results of the attitude angle is 0.27°.