三角網格錯誤偵測與特徵線萃取

Abstract

隨著三次元量測系統的進步，使逆向工程的應用逐漸受到重視；而曲面實體模型的重建是逆向工程中的重要步驟之一，尤其是作為輸入的三角網格檔案，其品質的好壞，以及鋪面時的處理步驟，將會影響曲面重建的效率；所以對網格進行錯誤與孔洞偵測並進行修正，與萃取特徵線為基礎的網格分割，對鋪面的過程有正面的幫助。 本研究以CAD與逆向工程廣泛所使用的STL作為處理對象，利用半邊結構建立網格間的拓撲結構，並運用AVL樹加快拓撲搜尋與後續處理的速度；針對網格模型常發生的多種錯誤形式進行整理，利用網格拓撲關聯性進行錯誤偵測，並提出錯誤修正與孔洞修補的方法；最後對網格進行特徵線萃取，以網格曲率變化為主，建立不同的特徵運算子與特徵門檻，使得到的特徵能符合使用者的需求，並依其進行網格分割，加快逆向曲面前處理的效率。

With the advance of 3D measurement system, the application of reversed engineering pays gradually much attention. Reconstruction of surface model is one of the most important issues in reversed engineering especially on the improvement of mesh file. The procedure of reconstruct surface will affect the efficiency and quality of surface. So the errors diagnosis, hole fill and feature lines extraction are the great benefit to later mesh processing. STL file is more popular input file format of CAD and reversed engineering and regards as the researching target. This research based on half-edge data structure for building the topological information of triangular mesh. Then, search algorithms based on AVL tree reconstruct efficiently the topological structure. We use the completed topological information of triangular mesh to detect errors and holes, then use algorithms based on edge's topological relation to correct and fill them. Finally, feature operator and threshold is needed to extract features with large curvature in the triangular mesh. Providing a flexible requirement to extract features and dividing mesh based on the feature line.