一种行星车视觉导航系统的立体校正算法

A Stereo Rectification Algorithm for Planetary Rover Visual Navigation System

针对行星车软着陆过程中,视觉导航系统参数变化后星面再标定难度大的问题,在传统立体校正算法基础上提出了一种适用于行星车视觉导航的弱标定立体校正算法。算法利用地面标定结果和对应匹配点进行优化计算,约束校正矩阵形式,保证了校正结果为以光心为投影点的单应变换,并进一步简化参数,降低计算复杂度。在理论上证明了算法的正确性,实验验证了算法的有效性。实验结果表明,算法校正结果与已知相机内、外参数的欧几里德校正方法性能相当,校正误差和校正畸变均较小,满足行星车视觉导航系统要求。

During the process of soft-landing for a planetary rover, the parameters of the video navigation system will be changed. There exists the problem about the increased difficulty for the re-calibration of the two cameras. Based on the common stereo rectification algorithm, a weak calibrated stereo rectification algorithm for the visual navigation system of the planetary rover is proposed in this paper. In this algorithm, the rectification matrix is calculated by corresponding points and parameters calibrated on the Earth; the rectification matrix is constrained to ensure that the projection is a homograph transformation whose projection point is the optical center; the computational complexity is reduced by reducing the number of the parameters. The rectification performance comparison experiments between our algorithm and the Euclidean Rectification algorithm have demonstrated the effectiveness of our proposed algorithm. The correction error and the correction distortion are both smaller, which meets the requirements about the vision system of the planetary rover.