本研究基於「分割-合併-成型」演算法之特性,以半自動方式重建房屋之三維模型,所利用之資料為多重疊數位航空照片。由於「分割-合併-成型」演算法可以接受不完整之屋緣線段,因此可以利用影像特徵萃取與影像匹配技術,求得未被遮蔽屋緣之三維線段,進而重建房屋三維幾何模型。首先,操作者可以人工點選房屋屋角之近似位置,據以進行局部的線型特徵萃取,再搭配特徵線段匹配與前方交會,以獲得屋緣之三維線段,進而使用「分割-合併-成型」演算法重建房屋之三維模型。實驗結果顯示,對於特徵明顯之線型房屋邊界,且高程為連續之房屋,本文所提出之方法可以有效地重建房屋之三維幾何模型。
This investigation based on the characteristics of ”Split-Merge-Shape” algorithm, utilizes a semi-automated strategy for building reconstruction. The original data used are the multi-overlapping aerial images. Since the SMS algorithm adopts incomplete roof-edges for building reconstruction, we proposed an interactive scheme that utilizes the image feature extraction and image matching techniques for non-hidden roof-edges extraction. First, an operator may manually locate the rough position of building corners in an image to obtain its approximate coordinates. Subsequently, the linear feature extraction and matching techniques were performed to extract precise location of roof-edges. Then, the 3D roof-edges can be calculated by means of photogrammetric space resection. Finally, we can adopt the SMS algorithm for building reconstruction. Experiment results show that the proposed scheme is effective provided that the building roof-edges are linear and high contrast with continuous height variation.