Building facade modeling using a RANSAC-style algorithm | |
| Author | Kumpee Teeravech |
| Call Number | AIT Diss. no.RS-14-05 |
| Subject(s) | Building--Computer simulation Three-dimensional display systems |
| Note | A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Remote Sensing and Geographic Information Systems |
| Publisher | Asian Institute of Technology |
| Series Statement | Dissertation ; no. RS-14-05 |
| Abstract | An overall objective of this dissertation is to propose a technique for generating LOD3 build-ing model. In order to achieve this goal, an algorithm for discovering repetitive pattern ona facade image is proposed. Thus, we propose an algorithm forgenerating building facademodel by automatically discovering the repetitive patterns of the dominant structures, suchas windows and balconies. The proposed algorithm follows a histogram-based approachthat analyzes the accumulated horizontal and vertical histogram profiles of image evidence,i.e., image gradients. In this dissertation, a histogram profile is viewed as a series of noisywave cycles, where a wave cycle represents the approximatedposition and dimensions of awindow. In that respect, the dominant frequency of the histogram profile should be highlyrelated to the repetitive pattern on building facade. Thus,the repetitive pattern is discoveredby iteratively fitting candidate sine waves to the histogramprofile using a RANSAC-styledalgorithm. After the best fit sine wave is obtained, the splitting lines, that are used to decom-pose the facade image into floors and tiles, are positioned inthe valleys of the resultant wave.We evaluate our algorithm using the publicly available facade image database. Additionally,our measurement dataset is also used. Comparisons between the proposed algorithm andseveral baseline techniques are also evaluated and discussed. The recovered repetitive pat-tern is then used to decomposed facade image of a building into tiles. After that, outline ofa window on each tile is estimated using a gradient-based technique. Finally, a 3D build-ing model is generated using an international standard, named CityGML. According to theresults, the proposed algorithms can discover not only regular pattern, but also irregularpattern.iii |
| Year | 2014 |
| Corresponding Series Added Entry | Asian Institute of Technology. Dissertation ; no. RS-14-05 |
| Type | Dissertation |
| School | School of Engineering and Technology (SET) |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Remote Sensing (RS) |
| Chairperson(s) | Nagai, Masahiko ;Honda, Kiyoshi (Co-Chairperson) |
| Examination Committee(s) | Dailey, Matthew ;Apichon Witayangkurn |
| Scholarship Donor(s) | Government of Japan |
| Degree | Thesis (Ph. D.) - Asian Institute of Technology, 2014 |