The Uncertainty Analysis Of Camera Calibration Parameters And Three-dimensional Measurement

Camera calibration is a critical step in recovering the three dimensional spatial information from two-dimensional image, it is also a very important research specialty in computer vision. The accuracy,reliability of the camera parameters is directly related to many subsequent applications, such as 3D reconstruction and measurement. Therefore, the analysis of the performance of the camera calibration parameters have important theoretical and practical value in the field of computer vision. In addition, in the measurement of computer vision, for specific using of the camera, prior to analyze and master the model of the camera and the measurement accuracy is very necessary and important, it will directly affect the final measurement accuracy, if the model of the camera is unreasonable.This paper focuses on the uncertainty of the camera calibration parameters and the selection of the camera model. The specific research contents and results of this paper are as follows:1. the probabilistic method is used for researching the uncertainty of camera calibration parameters and its influencing factors. Firstly, the formal description of camera calibration problem is given, then derived the analytical representation of uncertainty of the camera calibration parameters according to the error diffusion theory. According to the characteristics of the camera calibration problem, based on the structure of Jacobi matrix, which is deduced from the uncertainty of the camera calibration parameters, proving the relationship between the number of feature points of model plane on 3D special plane and the number of images and the uncertainty of the camera calibration parameters in theoretically, which is the main innovations in this paper. Finally, taken the Brown model for an example, the plane template calibration method is used for experimental verification. Experimental results show that, the analysis and derivation of the uncertainty of camera parameters is reasonable,credible. And to a certain extent, the increasing of the number of feature points is more effectively than increasing the number of images.2. For visual measurement engineering application background, the three-dimensional measurement precision of experiment are compared, which is equipped of Nikon50 of prime lens of Nikon D300S cameras, and select the reasonable imaging model. Experiment, the calibration of camera intrinsic parameters are still using the calibration method based on planar template, and the calculation of camera external parameters is based on the essence matrix decomposition method. For 8 different camera models, with different distance scene, different location of the feature points in the planar template, to reconstruct the Euclidean distance between feature points and straight Angle, and the reconstruction precision is compared, the experimental results show that the model for Nikon D300S cameras, lenses for 50 mm, take perspective imaging model of radial distortion coefficient of the reconstruction accuracy than other imaging model.