利用標準模型之形變及實體輪廓套合進 行物體三維模型之重建

3D Reconstruction Based on Model Deformation and Apparent Contours

蔣泳洋(Yung-Yang Chiang)

指導教授 ： 林惠勇

國立中正大學/工學院/電機工程學系/碩士(2013年)

摘要

本篇提出一套模型形變系統，主要用於調整物體模型以提升模型建製的品質。此系統應用於主動與被動式模型建製，主要目的是解決被動式方法模型品質不佳的問題。我們的方法是根據主動式建製的精緻模型為基礎，被動式建製的粗糙模型為約束條件，以此基礎與條件進行模型形變。模型建製則是以可見外殼為基礎，該方法使用容易，但對相機參數相當敏感，導致模型的細節與紋理不易體現。為了解決上述問題，我們提出模型形變的方法，目的是結合主被動法的模型品質與結構。在模型形變的過程，我們使用資料分群法將模型分成數個子模型，再以迭代最近點演算法調整這些子模型，使調整後的模型與目標物體逐漸匹配。最後在實驗中設計了一些例子，使建製模型的品質較差，但經由本篇方法的處理後，還是能適當地還原出物體的樣貌，從中驗證本篇方法的可行性。

關鍵字

模型形變

並列摘要

In this thesis, we present a model deformation system to enhance the quality of 3D model reconstruction. This system is used in active and passive model reconstruction, and the main purpose is to solve the general problem of poor quality for the passive method. Our approach is based on the exquisite model from active method with the constraints based on rough model from the passive method, in which the model deformation is carried on the basis and conditions. The model reconstruction is based on Visual Hull, which is easy to use but quite sensitive to camera parameters. To solve the above problems , we propose the method of model deformation to combines active and passive methods with the quality and structure. In the procession, we use the clustering method to divide the model into several sub-models first, then adjust these sub-models by the iterative closest point algorithm, let the model and the object be matched after adjustment. Finally, we designed the experiments in the case of poor quality to the models. After the adjustment by our method, we appropriately recovered the appearance of object, which verifies the feasibility of our method.

並列關鍵字

Model Deformation

參考文獻

[1] Ulrich Linsenmaier, Clemens Rock, Ekkehard Euler, Stefan Wirth, Roland Brandl, Dorothea Kotsianos, Wolf Mutschler, and Klaus Jürgen Pfeifer, ``Threedimensional

ct with a modified c-arm image intensifier: Feasibility1'', Radiology, vol. 224, no. 1, pp. 286--292, 2002.

[2] Baumgart and Bruce Guenther, ``Geometric modeling for computer vision'', Tech. Rep., DTIC Document, 1974.

[3] Yasutaka Furukawa and Jean Ponce, ``Accurate, dense, and robust multiview stereopsis'', Pattern Analysis and Machine Intelligence, IEEE Transactions on vol.32, no. 8, pp. 1362--1376, 2010.

[9] Carlos Hernández Esteban and Francis Schmitt, ``Silhouette and stereo fusion for 3d object modeling'', Computer Vision and Image Understanding, vol. 96, no.3, pp. 367--392, 2004.