一种基于非参数模型的相机内参校准方法

A Method of Camera Internal Calibration Based on Nonparametric Model

摘要:

影响大空间视觉三维坐标测量不确定度的主要因素有相机内参数、外部方位参数和特征成像质量。传统相机内参数校准方法需要参数模型化及全局最优化求解方法，会造成各内参数相关性过高，局部校准误差较大，对基于测角的测量方式不确定度影响较大、结合相机成像原理及垂线法，提出了一种基于物理参数校准的方法，对主点位置偏移、全视场范围非均匀镜头畸变等参数进行了精细校准。最后，以交比不变误差作为相机校准精度的评价方法，与传统校准方法的对比实验表明，该校准方法能够有效地提高测量不确定度指标，是一种简单、实用的校准方法。

Abstract:

The main factors affecting the measurement uncertainty of large space visual 3D coordinate include camera parameters, external orientation and feature imaging quality. Traditional camera internal calibration method needs parameter modeling and global optimization method, which causes a high correlation of various parameters and bigger local calibration error, greatly affecting the uncertainty of angle measurement. In this paper, combining the principles of camera imaging and the method of vertical, it proposes a calibration method based on physical parameters, thus fine adjustment is made on the parameters such as position offset of main point and nonuniform lens distortion of the whole field. Taking the cross invariance error as the evaluation method of camera calibration accuracy, compared with traditional calibration methods, the proposed method can effectively improve the indicators of uncertainty measurement.