| Summary: | 碩士 === 國立成功大學 === 測量及空間資訊學系碩博士班 === 100 === Three-dimensional building model is one of the major components of a cyber-city and is vital for the realization of 3D GIS application such as urban development and planning, real estate management, change analysis, and so on. Therefore, how to improve the efficiency and accuracy of reconstructed building model is an important research issue. This paper presents a data-driven method for semi-automatic reconstruction of building models from 3D structural lines of a building, i.e. their eaves and ridges. The imagery were acquired by an airborne multiple cameras imaging system (AMCIS), which contains one vertical and four oblique DSLR cameras. The tilted cameras have an off-nadir viewing angles of about 45 degrees. The collected images were calibrated and triangulated to obtain accurate interior and exterior orientation parameters. During building model reconstruction, we detect line segment first and perform multiple lines matching manually. Then, a space intersection algorithm based on line-plane intersection is applied to obtain 3D roof structural lines. Finally, we utilize the TIN-Merging and Reshaping (TMR), which is a line-based roof model reconstruction algorithm, to achieve the goal of semi-automatic building modeling. In this study, during line segment detection, we utilize an open source called Line Segment Detector (LSD) that a sub-pixel accuracy could be obtained. The results can describe the outlines of roof boundary accurately and reliably. After manual multi-line matching the ground coordinates of a roof structural line was adjusted using a total least squares method which utilizes singular value decomposition (SVD) to avoid singularity issue. For performance evaluation, several case studies including different types of roofs, such as flat roofs, gable roofs, and cambered roofs, will be discussed. Experimental results indicate that except for circular or curved roofs or edges, the proposed method can successfully and accurately reconstruct 3D building models. It is expected that the proposed method can be used in stereo-plotting instead of conventional point-based approach. Meanwhile, for the purpose of increasing the efficiency of three dimensional building modeling the proposed method can be extended as a fully automatic multi-line matching procedure in the future.
|
|---|
