论文部分内容阅读
针对无人机在连续飞行过程中的姿态求取问题,提出一种基于单目视觉的微型无人机姿态算法。基于无人机摄像机获得序列图像,利用图像尺度不变特性变换获取特征点信息,结合对极几何约束关系,运用随机采样一致性原理求解载体位姿变换信息,从而获得载体的导航信息。实验结果表明,通过单目序列图像获得的姿态角度变化精度优于0.1°,在180°旋转情况下的误差累加值小于1°。
In order to solve the attitude problem of UAV during continuous flight, this paper proposes a UAV pose algorithm based on monocular vision. Based on the sequence images acquired by UAV camera, feature point information is obtained by invariant feature transform of image scale, and the information of pose and pose transform is obtained by using the principle of stochastic sampling consistency in combination with the pole geometry constraint, so as to obtain the navigation information of the carrier. The experimental results show that the accuracy of attitude change obtained by monocular sequence images is better than 0.1 °, and the error accumulated value under 180 ° rotation is less than 1 °.