摘要
相机参数的准确性决定后续视觉任务的实现结果,为了提高相机校准的准确性,提出一种基于概率分布异常值去除和信息增益的内参确定方法。首先用一个摄像机对某个确定规格的棋盘图获取一组图片集,从图片集中随机选取一定数量的图片,使用张氏标定法做N组标定实验;然后将标定结果集按照其服从的概率分布进行异常值去除,将剩余的数据集中各个内参因素值分为3类,并且选择当前根节点下信息增益最大的参数类为最优节点,即以信息增益最大为主线选择最优路线;最后通过对最优路线中标定精度较高的结果集使用中值来估计相机参数。实验结果证明该方法与原始方法比较,提高了准确性。
The accuracy of the camera parameter determines the realization result of subsequent vision tasks.An internal parameter determination method based on probability distribution outliers removal and information gain is proposed to improve the calibration accuracy of camera.A camera is used to shoot a set of pictures of a certain size of the checkerboard,then a number of pictures are randomly selected in the picture set,and the Zhang′s calibration method is adopted to carry out the N-group calibration experiment for the pictures.The outliers are removed from the calibrated result set according to its probability distribution.The factor values of each internal parameter in residual dataset are divided into three categories.The parameter category with highest information gain under current root node is selected as the optimal node,that is,the optimal route has the maximum information gain.The camera parameters are estimated by using the median value of the result set with higher accuracy in the optimal route.The experimental results show that,in comparison with the original method,the accuracy of the proposed method is improved.
作者
刘文庆
冯锋
LIU Wenqing;FENG Feng(Ningxia University,Yinchuan 750021,China)
出处
《现代电子技术》
北大核心
2019年第5期48-52,共5页
Modern Electronics Technique
基金
国家自然科学基金资助项目(71561023)
宁夏大学创新项目基金(GIP2017049)~~
关键词
相机校准
概率分布
3D重建
相机内参
信息增益
最优路线
camera calibration
probability distribution
3D reconstruction
camera internal parameter
information gain
optimal route
