以骨架為基礎的三維特徵與模型檢索

Skeleton Based Feature Representation and Matching for 3D Model Retrieval

秦壽崙 ； 杜維昌

《南臺灣資訊科技與應用研討會》 2007年 (2007/05) Pp. 31-40

https://doi.org/10.29684/NANTAIWANZ.200705.0031

摘要

由於建構工具的日益成熟，使得製作三維模型的效率有很大的提升，但面對與日俱增的資料，三維模型檢索的相關技術自然成為目前熱門的研究議題。由於骨架是三維模型本質上的特徵，不易受形變因素所影響，亦與人類判斷物體外觀十分相符。在此一研究中，我們將探討以三維模型骨架為基礎的特徵擷取與模型檢索，並且在滿足不變的拓樸性質下對於非剛體模型的運動形變仍能保持高度穩定的檢索結果。在本文中，首先將三角月所組合而成的三維模型轉換成骨架表示方法，亦即透過對三維模型進行體素化、計算關鍵點、定義向量場等工作，讓關鍵點依照向量場方向連通起來而得到骨架。接著，使用多種統計直方圖作為幾何與拓樸性特徵的萃取方式，並籍此進行三維模型的有效檢索。我們實際採用普林斯頓大學三維模型資料庫為測試標的，以驗證其執行成效。

關鍵字

三維模型 ； 資料檢索 ； 三維骨架 ； 拓樸特徵

並列摘要

Retrieval by content of 3D models is becoming more and more important due to the effective development of rendering technology. The World Wide Web is enabling access to 3D models more and more by people. In this paper, we present a new method, referred to as Curve-Skeleton of 3D models, that develops on the topological feature approach, with adaptations to support effective retrieval by content. According to the method proposed, a skeleton is computed for each model, to obtain an invariant feature of deformable 3D models. The algorithm of skeletonization is based upon computing a repulsive force field over a discretization of the 3D models and using topological characteristics of the defined vector field. Then, instead of original 3D model, we support topology and extract invariant feature from the deformable 3D skeleton. Experimental results are presented, showing the effectiveness of the Curve-Skeleton method retrieval, also in comparison with other methods, and its sensitivity to model deformations.

並列關鍵字

3D model ； data retrieval ； 3D skeleton ； topological feature