具視點相依多層次精細度的地形顯像技術

Abstract

地形顯像系統中，多層次架構佔了很重要的地位。由於地形幾何資料量龐大，幾何顯像時，勢必要減少幾何資料，以達成即時、穩定的顯像速度。本論文採用了視點相關的多層次精細模型建構技術，在地形即時顯像系統內，根據使用者視點位置，動態地產生適合層次的幾何資料。幾何資料是以地形幾何節點及其相鄰關係所建立的四元樹為主。在這個樹狀資料結構上，我們利用時間共效性的特性，建立了動態連結串列，連結依視點而定所構成多層次模型之節點，並建立這些節點所組成的三角形串列，以提供每個顯像畫面所須的多層次幾何資料。在即時執行顯像動作時，系統會依據視點差值及視點範圍內擷取的原則，動態地決定地形四元樹上動態連結串列的節點及三角形串列。由於利用了視點相依式的多層次精細技術，以及四元樹架構的時間共效性，我們可以適當地簡化幾何資料及有效地加速系統運算時間。

In a terrain rendering system, level of detail (LOD) modeling is an important technique, which is used to reduce the terrain's geometry complexity in order to achieve real-time frame rate. In this thesis, we develop a view-dependent LOD technique aiming to construct appropriate level of detail according to the eye point. The terrain data is organized as a quad-tree that implicitly records the adjacency relations among vertices. On the quad-tree, we maintain a dynamic linked list that links all activated vertices and an active triangle list for all triangle to be rendered. In each frame, system will traverse through the dynamic linked list and refines or simplifies the LOD model according to a screen error and view frustum culling test, and update the active triangle list accordingly. The proposed method exploits criteria for view-dependent LOD selection and frame coherence in order to achieve geometry reduction and efficiency speedup.