使用線段實體模型由多張物體輪廓重建3D物件之技術

Abstract

近年來，由於虛擬實境的盛行，將現實生活中實際存在的物體，以自動化的方式建構其三維尺寸幾何模型，已成必然的趨勢。本篇論文提出一套以線段模型建構三維物體的方法。該方法利用校正過的相機對物體拍攝多方位的影像，便能推論出物體的三維幾何尺寸模型。 我們的方法，是採用體積交集法(Volume Intersection)的方式，使用拍攝相片中的物體輪廓，界定出線段模型中屬於物體的部分。且由於模型中線段的解析度可動態調整，故能主動逼近物體外緣，以較少的線段獲得與高密度線段品質相近的結果。最後並能將線段模型轉化為物體的多邊形網格模型，方便網路展示或是提供給其他三維圖學工具應用。 我們提出的方法，與傳統體積交集法中的八分樹(Octree)建構法比較，具有處理時間快且不需龐大儲存空間的優點。

Because of popularity of virtual reality (VR) applications in recent years, automatic reconstruction of 3D geometry models from real-world objects has become a very important issue. This paper presents a method to reconstruct 3D real objects using line-based solid model from a set of images captured with a calibrated camera at different viewpoints. Our method is to use the volume intersection approach to derive the line-based solid model from multiple silhouettes of the object. Since the resolution of lines in the model can be adjusted dynamically, the model can describe the boundary of the object more precisely without increasing the overall resolution. With this property, we can use few lines than the high resolution model, while maintaining a similar quality. Finally, we convert the reconstructed object model to a 3D triangular mesh. The mesh can then be browsed over the internet or used by other computer graphic applications. In comparison with the traditional octree reconstruction method, our method takes less processing time, and does not require a large storage space.