摘要
针对用于大空间运动分析的运动参数光电探测系统中多台高速摄像机的内外参数校准需求,制作了一种两端及中间各安装一个红外反光标志球且两端球距为已知(精确校准)的刚性球杆。将这种特制的刚性球杆在测量空间内多次随意移动,并用多台前端装有红外发光板和滤光片的高速摄像机摄取其图像,即可实现焦距不同的多台摄像机的内外参数校准。假定主点位于像面中心附近的某个位置,通过Hartley改进8点法求出基本矩阵,利用极点和外极线约束线性地求出各台摄像机的焦距。接着,求出本质矩阵,对其进行奇异值分解后得到旋转矩阵和比例因子意义下的平移向量。利用三角法确定刚性球杆两端点的重建坐标和距离,与标准距离对比确定比例因子。最后,通过评价函数将摄像机校准转换成寻找摄像机最佳主点配对的非线性最小化问题,运用改进的模拟退火进化策略迭代求解出最佳主点配对,继而求解出摄像机的其它内外参数。标定实验表明,焦距和主点求解精度达到了0.1pixel。对比测量实验表明,对长为750.607mm的刚性杆进行长度测量,标准差达到了0.046mm。与传统方法相比,本方法扩展到了多台摄像机应用场合并允许摄像机焦距各不相同,不需要对球杆的运动做任何限制,能够同时求解出摄像机内外参数,而改进的模拟退火进化策略改善了算法收敛速度和全局收敛性能。
According to the requirements of calibrating multi-cameras in motion parameter optoelectronic measurement systems used in big space motion analysis, a novel approach of calibrating the internal and external parameters of multi-cameras is proposed. Firstly, a principal point is assumed at a certain location around the image plane center, and the fundamental matrix is found out by Hartley＇s eight-point algorithm. Then, different focal lengths of different cameras are obtained linearly through a epipolar geometry and it＇s constraint. After educing the essential matrix, the singular-value-decomposition is carried out to obtain the rotation matrix and translation vector with a scale factor. Moreover, the coordinates and the distances of the ball bar are computed by a triangulation method, and the scale factor is found out through the true distance of the ball bar. Finally, the estimation of internal and external parameters of the multi cameras is transformed to a nonlinear minimization problem by a cost function, and the optimal principal points of two cameras are found out iteratively by using the improved simulated annealing evolution strategy. The experimental results show that the precision of focal length and principal point is up to 0.1 pixel and the stardard deviation bar length is 0. 046 mm. Compared with traditional methods, the novel calibration approach expands the applications of the multi cameras and can find out the internal and external parameters simultaneously without the constraint for the ball bar movement. These results prove that the improved simulated annealing evolution strategy accelerates the convergence speed and improves the global convergence performance.
出处
《光学精密工程》
EI
CAS
CSCD
北大核心
2009年第8期1942-1952,共11页
Optics and Precision Engineering
基金
国家自然科学基金资助项目(No.50275040)
关键词
摄像机校准
球杆
自由运动
模拟退火
进化策略
camera calibration
ball bar
free motion
simulated annealing
evolution strategy
