| Coming in with the computer, it has been a hot spot, in computer vision and digital photogrammetry, which obtaining the information of geometry and texture of the shot objects by the technologies of computer image processing and projective geometry. And camera calibration is one of the most important contents.It can be said that research of the theory of camera calibration has been sophisticated. And lots of effective approaches had also been presented by the experts in computer vision and photogrammetry. However, none of them can be used in all applications and all environments. Ordinarily, they can not satisfy both parties in precision and practicality. In respect of high precision approaches, they usually need 3D highprecision control field, especially 3D highprecision indoor control fields. And in respect of practical approaches, their precisions may be low. So how to obtain highprecision camera parameters with simple reference objects and reliable approaches is still the first most problem in the research field of camera calibration. This paper mainly aims at the following problems: reference object and non-parameters distortion models. Among them, the method based on Completely Flat Liquid Crystal Display (LCD) is successfully proved to be an excellent calibration method by testing with lots of analog and actual data, which is presented by this paper.As a latemodel peripheral for the computer, LCD has been popular in many families. Moreover, because of the sophisticated manufacture technology, the geometrical precision and visual effect of LCD have been excellent too. Therefore, this paper introduces LCD into the camera calibration in the base of the calibration method based on the planar pattern, So as to replace other highprecision planar patterns or 3D reference objects. Because of the flexibility to draw grid points on the LCD with special program and create 3D control field by combining two LCDs together, calibration based on single LCD or double-LCDs can fully conform with the calibration's demand of convenience and practicability. This can help us to make out a highprecision and practical camera calibration system.Commonly, normal calibration approaches only consider functional distortion models, especially radial and decentering distortion (named as parameter distortion model too). It is enough for more surveying with middle precision. However, because of the complexity of the manufacture technology of digital camera, it is considerable that whether the simple parameter distortions can completely fit with the imaging distortion of the digital camera. For example, those errors caused by combination of CCD units or electricity function may not be ignored, which is more important to the great highprecision surveying. Moreover, some uncertain errors may exist in the course of the surveying, which do not belong to the camera. So we also need to research on some other non-functional distortion models, such as interpolation or hybrid models. According to the existent reference documentations, this paper presents an improved interpolation distortion model based on finite elements model. And in the other hand, this paper also tries to use the Posteriori Compensation (PC) method for camera calibration, which is usually used for systematic errors compensation in the course of self-calibration bundle adjustment of photogrammetry. Lots of experiments show that both of them can help to obtain good results. Especially, the combination with PC and parameter distortion model can not only help to improve the calibration precision, but also be carried into execution easily. Experiments show that it is more flexible and convenience than the approaches presented by Munjy,Luhmann and Hastedt, which are based on the finite elements model.As a novel reference object for the camera calibration, how to obtain high-quality images for calibration and Non-planarity analysis of LCD are two most key problems for the research. In the first problem, we presented a new and efficient shooting method after lots of experiments. In the second problem, we got some qualitative and quantitative results by the method of stepwise regression analysis. And then we can further improve the calibration precision when Non-planarity of the LCD is corrected with the analysis results.According to the all research contents, this paper makes out a highprecision and practical software system that integrates several calibration methods.In order to further prove the efficiency of the methods, this paper, finally, presents a new method of close-range 3D reconstruction with rotational shooting and multi-baseline stereo matching.
|
PDF Full Text Request