A Method For Camera Calibration Based On Cross-ratio Invariability

In recent years, with the continuous development and the needs of human science and technology, computer vision (machine Vision) is in-depth study. As camera calibration is the first step in computer vision ,which gradually become a hot topic in vision research.Camera calibration has an extensive use in the many image applications fields such as virtual real -ity,photogrammetry,medical diagnosis,industrial inspection,three dimensional reconstruct- ion and so on.With two-dimensional images, if the object can be accurently reconstructed by three-dimensions,camera must be accurately calibrate and for space motion analysis and mea- surement of precision instruments as well.Though the domestic and international scholars have done a large amount of positive research in this field,more effort should be paid to imp- rove the precision of camera calibration.Corner detection of image plays a key role in the camera calibration, the precision of camera calibration has directly affected by the accuracy of corner detection. Inthis paper, camera calibration technique has systematically studied and summarized, especially image corner detection.Against of the limitations and the shortcomings of current corner detection algorithm, three parts are mainly developed in this paper:(1) Aiming at the problems of current corner detection algorithm, for instance corner information is easily lost, deviated and susceptible to noise. A dynamic method based on the principle of cross-ratio invariance for camera calibration is proposed.Corner not only can be well extracted and calibrated via cross-ratio invariance of perspective projection of space geometry transform, but also Non-corner is as well. Especially, this method can effectively overcome the defects of corner detection.(2) As high accuracy experimental template hardly well designed and is used to shot in specific environments(uniform illumination),it leads less practicability,the intersection line templatea is proposed,which is easily designed and used to shot in natural environments.In order to accurately match all line with any intersection,A method for matching straight line based on geometric invariance is proposed.which is the approach that is not based on easily miss and deformed gray,length and angle of line support region and so on.The experiments showed that the algorithm has great robustness,especially to rotated image and line which is partially sheltered.(3) With the largest computer vision library OpenCV,this paper implements a display interface of the intrinsic parameters of the calibration via Visual C + +6.0 platform.this work will be prepare for more following experimen.