The image based modelling of Jinnamgwan

Abstract

Jinnamgwan is a National Treasure of Korea and a symbol of Yeosu city. It's a single-story wooden building which was built in 1599 after the Imjin War. Jinnamgwan has 68 high columns which is supporting the collar beams. The side blocks of the building have two side beams whose heads are stretching above the girders. Those many columns make it more difficult to perform image-based modeling using unstructured image set. Many researchers have combined both photogrammetry and terrestrial laser scanning in different ways. In our paper, we used only images which can be acquired from some internet sites. Structure from motion refers to the process of estimating three-dimensional structures from two-dimensional image sequences. It has been studied in the fields of computer vision and visual perception. We applied a modified structure from motion algorithm for its 3D modeling using the Photoscan software. Close range images were used for column modeling and aerial images were used for roof modeling. A tens of multi-temporal close range images were acquired from web sites such as google, some Korea portals etc.. Because we could not construct a roof model from a few close range images of roofs, three digital aerial images were processed. In our experiment a building model of Jinnamgwan could not be made using all images at the same time. So all close-range images were divided into sub-class groups representing front and side facades. The model from each group's images constitutes a part of Jinnamgwan. In the future, to integrate sub-models into a whole wall model accurately, we will do a simultaneous adjustment to compensate for a systematic errors applying a 3 dimensional conformal transformation to point clouds of sub-models.