| 3D modeling is an important subject in the research of Computer Vision, which is widely used in many areas such as VR (virtual reality), antique protection, film industry, manufacture industry, CAD, CAM and clothing industry etc. Among the methods of 3D modeling, the method based on images is a popular and promising subject, and it is in great need at present because it can get accurate 3D data from an object cheaply without contacting it. Consequently, it is significant to study on this method in theory and practice.The recent techniques of 3D modeling are analyzed and the method of 3D modeling based on images is studied in this paper, including the system principle, system calibration, image raster extracting and indexing, and 3D data integration etc. The main work of this paper is as follows:1. The method of integral calibration is studied in this paper. By this method, camera intrinsic parameters, baseline length, and the angle between the raster and the baseline need not be calibrated alone. Once each light plane equation in the world coordinates system is to hand and the camera is calibrated, the system is calibrated wholly with high precision.2. The corner-matching optimization method of corner detection is studied. This method greatly improves the operation speed and reduces the cost of corner-matching, while the precision remains high.3. A new raster pattern and raster indexing method of the new pattern are studied. By this method, the indexing speed is improved compared with traditional method of dynamic programming algorithm. Besides, this method possesses high correct rate even if the raster is sheltered.4. An easy and practical method of measuring angle is presented, which is measuring the angle of revolving turntable by solving vanishing point. The experimental results show this method possesses high precision with less accurate measurements, and is easy to be implemented.5. Post-processing of 3D data is studied. As a result, the data we get can be... |
PDF Full Text Request