基于一维标定物和改进进化策略的相机标定

为了解决运动参数光电探测过程中的相机标定问题,制作一种两端及中间各安装一个红外反光标记球的一维标定物.不需要其它复杂标定装置,只要将这种特制的一维标定物在测量空间内多次随意移动并摄取其图像,即可实现标定.算法首先假定主点位于像面中心附近的某个位置,再求出焦距、旋转矩阵、平移向量和比例因子,最后通过评价函数将相机标定转换成寻找两相机最佳主点对的非线性最小化问题.在传统进化策略中引进个体的自我改进系数、个体间距离等概念,提出了求取子代个体间的欧式距离并排序的方法,设计了搜索最佳主点对的改进型进化策略算法.与传统标定方法相比,基于一维标定物的方法克服了多相机场合的遮挡问题,改进进化策略的引入打破了一维标定物需做某种特殊运动的限制,使一维标定物自由运动时相机内、外参数的同时求解成为可能,改进的模拟退火进化策略改善了算法的全局收敛性能并加快了收敛速度.

Cameras Calibration Based on One-Dimensional Reference Object and Improved Evolutionary Strategy

To calibrate the intrinsic and extrinsic parameters of cameras in motion parameters photoelectric measurement system, a novel approach based on improved evolutionary strategy and surveying a rigid one-dimensional object carrying three markers on its middle and extremities moved inside the working volume, which does not require other complex calibration equipments, was proposed in this paper. Firstly, it is assumed that the principal point is located at a certain location around the image plane center. Then the focal length of cameras, the rotation matrix, translation vector and the scale factor are obtained. Finally, the estimation of intrinsic and extrinsic parameters of cameras is formulated as a nonlinear minimization problem of a cost function. The simulated annealing evolutionary strategy algorithm was improved by introducing the self-improved factor and the Euclidean distance between individual. Consequently the global convergence performance of the search algorithm of this method was mended and accelerated. By introducing the evolutionary strategy, the new method can find out the intrinsic and extrinsic parameters all together, and need not constrain the movement of the one-dimensional object in comparison with traditional methods.