Research On Digital Camera Calibration And Related Techenique

Human vision can be regarded as a complicated process from feeling to consciousness. It also means that 3-d world projects 2-d image, as well as cognitive process of the content and the meanings in 3-d world according to the 2-d image. With the advent of signal processing theory and computer, computer vision, a emerging discipline , arose. People complete the process of visual information acquisition and transmission, obtaining environmental images and converting them into digital signals by camera. People also implement the processing, storage and understanding of visual information by computer. The process of restoration of 3-d objects visible surface geometric structure from 2-d image called 3-d reconstruction. With the rapid development of the computer hardware, software, image acquisition and processing technology, the theory and technology of 3-d reconstruction has been widely applied in medical image processing, robotics, character recognition, industrial inspection, military reconnaissance, geographical mapping, field surveys and virtual plant visualization, etc.Camera calibration is an indispensable step in 3-d reconstruction, including the internal structure of camera parameters such as the coordinate of main point and the external parameters such as the rotation of the camera and translation. The cheap experimental equipment, simple experiment environment, fast calibration speed and precision of calibration are the directions of camera calibration research. Digital camera calibration is one of the research hotspots.This thesis summarizes the camera calibration technique comprehensively, including introduction to background, situation of study, research status of the camera calibration research, categories and rudiments of camera model. Two new calibration methods with simple processes are proposed, which is effective to improve the classic camera calibration. After several evaluation criterions are summarized systematically, a new evaluation method, weighted average magnitude of error, is proposed. The new ideas in this thesis are summed up as follows:(1) A calibration method is proposed, which calculate the internal parameters based on line. This method considers camera imaging model as an ideal model without distortion. According to the setting information of digital camera and geometric relationships in the camera imaging model, we obtained the internal parameters of the digital camera conveniently. And the result is used as the the initial numerical value for iteration in Zhang's method. The experiment proves that this method is available.(2) A new calibration method based on the 3-d object was introduced. This 3-d object consisted of three lines which were vertical and intersecting mutually in two mutually vertical and intersecting planes. This 3-d object can be manufactured conveniently, and low computational complexity. Experiments show that the calibration results of this method is rather accurate.(3) Based on the current four methods which evaluate camera calibration results, image noise, the number of images, comparing the experimental results with the classic methods, computation of the projective points or back-projection points errors,the method of weighted average magnitude of error was proposed,which evaluate the result of camera calibration.