摘要
针对双目视觉中两摄像机的固定约束关系,提出了基于双目视觉的光束法平差新算法。对传统光束法平差法方程构造过程进行了改进,使每次迭代过程中只需对拍摄的两幅图像中一个摄像机的外参数进行优化,更好地将两个摄像机绑定为一个摄像机,使优化精度更高;新算法中,法化矩阵的维数比传统算法降低了12,所以每次迭代计算所需要的时间也会减少,对于迭代次数越多的优化过程,新算法和传统算法计算所需时间差越明显。模拟实验和实际测量均表明,新算法比传统算法优化精度高,计算所需时间短。
In this paper,we propose a new algorithm for bundle adjustment based on the fixed constraints between the two cameras in stereo vision.The new algorithm improves the normal equation of traditional bundle adjustment,and it only needs to optimize extrinsic parameters of one camera in each iteration.The two cameras can be better bounded as one and it can achieve a higher accuracy.The dimension of the matrix in the new algorithm is half of that in the traditional algorithm.Therefore it will reduce the time of calculation in each iteration.Simulation and actual measurement experiments both prove that the accuracy of the new algorithm is higher than that of the traditional algorithm and the new algorithm costs shorter time.
出处
《光电子．激光》
EI
CAS
CSCD
北大核心
2011年第6期888-892,共5页
Journal of Optoelectronics·Laser
基金
国家自然科学基金资助项目(60838002)
关键词
双目视觉
数据拼接
光束法平差
误差方程
法方程
stereo vision
data registration
bundle adjustment
error equation
normal equation
