A Research Of Method For 3D Reconstruction Of Ancient Building

Currently, Cyber City constructing is a popular research project. Simultaneous data-collection and Virtual Reality imitation technology can be adopted to construct some regular buildings, covering a large area and multiple objects. However, exquisite ancient buildings require the careful reconstructing respectively. The difficulty of such a reconstructing work lies on the complex structure of the ancient building which becomes an obstacle for acquiring 3D space data of its each component and organizing the space data in constructing a model. This dissertation offers a systematic analysis on the process and methods for highly accurate 3D reconstructing from the perspective of digital sequence image, taking advantage of the rigorous theories concerning digital photogrammetry and some frontline achievements in computer vision.The accuracy of acquired 3D data exerts direct influence on the accuracy of constructing a model. The 3D data of different parts of an ancient building require different policies. Such regular parts as the wall, the pillar and the stylobate, ask for the abstraction of marginal lines to get the ancient building's profile margin. Then, the image coordinate for the profile line's end-points can be got based on these lines' intersection. And the 3D space coordinate describing regular parts' angle points can be acquired with the restriction of cor-plain conditon forward intercourse model. Whereas, the irregular parts such as the eave need adopting the method for construction TIN with restriction of border conditon which constructs irregular triangle net with the scattered feature points selected for describing the entity to present the surface of the ancient building. In addition, this dissertation mentions policies of how to select the feature points on the irregular surface of an ancient building. Calculating out the coordinate numbers for each feature point in a model needs interior orientation element parameter and relative orientation element parameter of its image. Interior orientation element parameter is got from Camera calibration, while relative orientation element of an image is from sequential relative orientationThe acquired coordinate of relative space is considered to organize its numbers so as to construct a model. Designing a set of document format for the model of a building, it is a convenient tool to organize the 3D data of feature points in all the building's parts. In the end, OpenGL function interface would combine the 3D data acquired of the model with their correspondent texture image data, producing the VR(Virtual Reality) model.