The Selection Of Camera Distortion Models And Distortion Correction

In the field of computer vision, the camera calibration problem has always been a very important research direction. Traditionally, camera calibration of the related research mainly focus on a given model parameter estimation problem. However, if the camera model selection is not reasonable, then no matter what method, it is difficult to achieve high precision calibration results. So, for a specific camera, how to select a reasonable imaging model is first need to solve in the camera calibration problem. In addition, the radial distortion is an important problem in digital image analysis, the traditional distortion correction method is mainly based on the calibration object in the scene, such as based on linear features in the scene or based on the known calibration template structure, etc. But this kind of method for the SFM (Structure from Motion) and other problems has certain limitations. So to explore corresponding points directly based on image distortion correction method has vital significance.This paper around the camera distortion model comparison and choice, the image distortion correction methods based on corresponding points in the research, the concrete research contents and results of this paper are as follows:1. This paper proposes a camera distortion model comparison and selection method. First analysis and define the three performance criterions of metric camera model, including:reconstruction precision and generalization performance of different models, model to estimate the stability and model parameter uncertainty. Then, by the lens distortion smaller standard lens, the distortion of medium short focal lens and the great fisheye lens distortion as an example, using the three performance criterions, for different distortion model have carried on the experiment analysis and comparison respectively. The results show that this paper proposed the performance criterions is reasonable and feasible, thereby provides an effective way to correctly select the camera model.2. Proposed a distortion correction method based on corresponding points of image. Its basic idea is that, first, according to the image distortion points establish a nonlinear constraint equation based on planar Homography, then to solve the equation adopting iterative method. The main characteristics of this method:the parameters and the distortion parameters of Homography adopt the successive iteration strategy in the process of iteration, which effectively improves the algorithm robustness. The experimental results verify the feasibility of the method. At the same time, the study found that the point positioning accuracy of the image corresponding has great influence to the calibration results.