標題: 應用電腦視覺技術實作以2D影像資訊建立3D建物模型於iOS平台之研究
The Application of Computer Vision Techniques for Image-based 3D Reconstruction on iOS Platform
作者: 劉廣德
Liu, Kuang-Te
趙文成
Jau, Wen-Chen
土木工程學系
關鍵字: 電腦視覺;影像處理;三維重建;電腦繪圖;BIM;iOS;iPhone;OpenCV;從運動求得結構;OpenGL;Computer Visualization;Image Processing;3D Reconstruction;Computer Graphics;BIM;iOS;iPhone;OpenCV;Structure from Motion;OpenGL
公開日期: 2011
摘要: 在電腦視覺及影像處理的領域當中,利用二維影像資訊來重構出三維虛擬模型 ,一直以來都是相當重要的研究主題,且伴隨著資訊技術的迅速發展,三維重建技術現今已被廣泛使用在電腦繪圖、工業檢測與人機互動等各個領域上。然而在我們的AEC/FM領域中,建立三維建物或場景模型的方式,往往還是通過傳統的電腦繪圖軟體(包含BIM軟體)或是選擇較為快速、精準但價格卻十分昂貴的三維雷射掃描儀來達到模型建置之目的,因此本論文的研究目標為透過一個相對方便、快速且低成本的方法來重建出建築物的三維模型。 在本研究中,以iOS4.3 – iPhone模擬器作為主要開發平台,使用iPhone上所搭載的數位相機來對目標物進行拍攝,從而取得二張或多張的視圖影像,然後利用Xcode Developer結合Intel OpenCV函式庫來對其進行圖像平滑、相機校準、形變校正、特徵擷取等影像處理演算法之計算,再利用從運動中求得結構的方式,得到各影像間的特徵點來進行對應,進而算出物體的三維坐標資訊,最後利用顯示卡繪圖函式庫OpenGL與OpenGL ES將其繪製成三維模型,且可透過旋轉、移動及縮放等方式觀看模型,來達到簡便、快速、擬真的二維影像三維模型重建之效果。
In image processing and computer visualization area, a 3D virtual model based on 2D image information is one of the main interests. With the rapid IT progress, 3D model reconstruction are now widely used in computer graphics, industrial QA, GUI, etc. However, in AEC/FM field, 3D model is constructed by conventional computer graphic software, including BIM or 3D laser scan. This study develops a relatively more convenient and lower cost method to re-construct the 3D model. An iPhone with iOS4.3 simulator was used as the platform for development. The camera of iPhone was used to take two or more photos of the object. Then Xcode developer and Intel OpenCV function library were used to smooth, calibrate, and determine characteristics of the photo. 3D coordinates were then obtained. OpenGL and OpenGL ES are used to plot the 3D model and the model can be viewed in different angles by rotating, shift, zooming. The new method is then verified by comparing the real structure and the reconstructed 3D model and good correlations are confirmed.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT079916584
http://hdl.handle.net/11536/49608
Appears in Collections:Thesis