以多張近景影像萃取牆面三維線段之研究

3-D Line Extraction for Building Façade Using Multiple Close-Range Images

張智安(Tee-Ann Teo) ； 高崇軒(Chung-Hsuan Kao)

《航測及遙測學刊》 16卷2期 (2011/06) Pp. 127-137

https://doi.org/10.6574/JPRS.2011.16(2).4

摘要

三維房屋模型可由簡單的矩形模型，逐漸以屋頂、牆面及室內結構加以細緻化，本研究目的為發展自動化牆面線形結構萃取程序，以提升房屋模型的細緻度。研究中結合多張近景影像及初始房屋模型重建牆面三維線段，主要工作包含自動化方位重建及牆面三維線段重建。方位重建使用加速強健特徵點（Speeded Up Robust Features, SURF）匹配產生影像間之共軛點，配合人工量測之少量控制點進行光束法平差，完成影像方位參數求解。三維線段重建中，首先萃取影像中牆面二維線形特徵，再以物空間匹配方式產生三維線段之點群，接著以隨機抽樣一致演算法（Random Sample Consensus, RANSAC）由三維點群計算三維線段參數，最後完成牆面三維線段重建。實驗成果顯示，自動化方位重建檢核點之均方根誤差在三軸方向分別為1.6公分、3.2公分及2.2公分；物空間匹配之均方根誤差在三軸方向分別是4.9公分、5.1公分及6.3公分；而線段重建之均方根誤差約為10公分。

關鍵字

多影像匹配 ； 物空間匹配 ； 線形特徵匹配

並列摘要

The detail of a building model can be distinguished into block models, roof structures, facade structures and indoor structures. The detailed model is not only similar to its true appearance, but also can be applied to more delicate aspects, which may facilitate the decision making procedure. In order to improve the level-of-detail of a building model, this research has developed an automatic line extraction procedure for facade, including orientation determination, multiple images matching and line fitting. A large number of conjugate points in multiple images are generated by Speeded Up Robust Features (SURF). Orientation determination can then be done by bundle adjustment using tie points and control points. Aiming on linear features, object-based matching is applied to decrease the effect of image difference caused by scale, rotation and relief displacement. Finally, 3D line fitting is done by Random Sample Consensus (RANSAC). The experimental results indicate that the Root-Mean-Square-Error (RMSE) of orientation determination in X, Y, and Z directions are 1.6cm, 3.2cm and 2.2cm, respectively. The RMSE of object-based matching in three directions are 4.9cm, 5.1cm and 6.3cm, respectively. Moreover, the RMSE of 3D line fitting is about 10 cm.

並列關鍵字

Multiple image matching ； Object-based matching ； linear feature