相机标定参数对双目视觉测距精度的影响分析

Analysis of the Influence of Camera Calibration Parameters on Ranging Accuracy of Binocular Camera

相机标定参数是影响标定结果的重要因子,在标定图像数、特征点数、标定区域3个参数的基础上考虑了面积占比与特征点点位分布两个新的参数,并设计单一变量实验进行了研究。结果表明,系统的测距误差随着面积占比的增加而减小且呈现出近似线性变化的趋势,集中型点位分布的标定板用于标定较狭长型分布使测距误差平均减小了39%。为分析各因子与测距误差之间的关联性,采用基于均匀实验设计的灰关联分析法对相机标定参数进行分析,定量获取了各参数与测距误差之间的灰关联度,结果显示5项参数灰关联度均大于0.6,与测距误差之间存在较强关联性,关联度最大的为标定图片数,其次为点位分布,最后为特征点数、视场面积占比、标定区域,三者相近。实验结论可为相机标定时参数的优先级考虑及取值提供一定指导并为后续相关研究的实验设计、数据分析提供方法参考。

Camera calibration parameters(CCPs)are key fac⁃tors affecting camera calibration results.Besides the three CCPs(the number of calibration images and detectable points,calibration regions),this paper takes two new CCPs(area proportion and point distribution)into consideration and performes single variable experiment to study them.The ex⁃perimental result indicates that ranging error will reduce by in⁃creasing area proportion and there is a trend of linear change between them.In addition,compared with scattered calibra⁃tion board,ranging error decreases by 39%by using concen⁃trated calibration board.To analyze the relation between CCPs and ranging error,this paper uses grey relational analy⁃sis based on uniform design method to quantitatively obtain the grey correlation degree between each parameter and rang⁃ing error.The grey correlation degrees of all CCPs are above 0.6,indicating that these five parameters have a lightly strong relation with ranging error,the biggest is the number of cali⁃bration images,the following is point distribution,the other three factors have approximate values.The conclusion men⁃tioned above can provide some suggestions for priority consid⁃eration and value choice of CCPs in camera calibration,also it can provide a method reference for experiment design and data analysis in subsequent studies.