摘要
由三维目标与其二维图像估计相对于目标的位置和姿态是计算机视觉成像中的一个重要问题。正交迭代(OI)算法是一种快速、且能全局收敛的姿态估计方法,但是当数据恶化时,不能给出正确的旋转矩阵。本文改进了该算法中旋转矩阵求解方法,避免了旋转矩阵求解中出现的错误。建立了模拟实验系统,使用改进的算法进行测量,在0.45～5.20m的范围内,摄像机到目标距离的相对误差小于±0.41%;在距离为3m时,旋转角度测量误差小于±1.8°。数学仿真结果表明,改进后算法的抗噪声能力得到改善,结果更为准确,且能快速收敛。
Camera location and pose estimation from the 3D object and its 2D image are important problems in computer vision. The orthogonal iteration（OI） algorithm is analyzed, the rotational matrix solution is refined. Experiment results show that, using the improved algorithm, the relative errors of distance from camera to object are less than ±0.41%, in the range of 0.45-5.20 m, and the rotation angle errors are less than ±1. 8° in a distance of 3 m. Simulation results show that the improved method can fast converge, rapidly, increase the noise resistance performance, and give more accurate rotation and translation parameters as compared with the OI algorithm when the data corrupted.
出处
《航空学报》
EI
CAS
CSCD
北大核心
2007年第4期943-947,共5页
Acta Aeronautica et Astronautica Sinica
基金
"十五"预研项目(41324020201)
关键词
近景摄影测量
视觉导航
计算机视觉
位姿测量
交会对接
close-range photogrammetry
vision navigation
computer vision
pose estimationr
rendezvous and docking
