摘要
提出了一种对高程度畸变图像的畸变纠正方法,并同时比较不同相机畸变模型纠正效果。由于无畸变的真值坐标不能在图像上直接量测,首先基于透视投影中的射影不变性得到相机在无畸变情况下的真值像素坐标。各畸变模型的最优参数通过最小化畸变模型纠正结果与无畸变真值坐标的差值来求解,模型参数的初值为最小二乘的线性闭合解或零,然后通过Levenberg-Marquardt(LM)算法迭代得到模型参数的最优值。各畸变模型的纠正效果通过棋盘中横向和纵向共线程度的量化统计指标来进行评价。同时通过比较畸变模型纠正结果和基于射影不变性计算的真值分析不同畸变模型对相机畸变的模型拟合能力。实验结果表明,本文提出的方法能够有效地对高程度畸变图像进行纠正,并比较和评价不同相机畸变模型的畸变纠正效果。
In this paper,a method for correcting distortion of seriously distorted images is proposed,and the correction effects of different camera distortion models are compared.Since undistorted coordinates cannot be directly measured on distorted image,the undistorted coordinates are obtained based on the projective invariance in the absence of distortion.The optimal parameters of different distortion models are solved by minimizing the difference between correction results of distortion models and the undistorted coordinates.The initial values of model parameters are least squares linear closed solution or zero,and then optimal values of model parameters are obtained by Levenberg-Marquardt(LM)algorithm.The corrective effects of different distortion models were evaluated by quantitative statistical indicators of horizontal and vertical collinearity of chessboard.The different model fitting ability is analyzed by comparing correction results of distortion models and the calculated true values based on projective invariance.The experimental results show that proposed method can effectively correct seriously distorted images and evaluate correction effect of different distortion models.
作者
邱歆煜
黄玉春
QIU Xinyu;HUANG Yuchun(School of Remote Sensing and Information Engineering,Wuhan University,Wuhan 430079,China)
出处
《测绘地理信息》
CSCD
2021年第S01期100-104,共5页
Journal of Geomatics
基金
国家自然科学基金(41671419,51208392)
武汉大学学科交叉项目(2042017kf0204)
关键词
畸变模型
交比
共线约束
误差函数
像素偏离
distortion model
cross ratio
collinear constraint
error function
pixel deviation
