網路虛擬城市提供使用者一個快速獲取真實世界資訊的工具。地形重建在虛擬城市中為一重要因子，當地形複雜度增加或是範圍增大時，常因資料量的增加而造成傳輸的延遲，因此有必要針對此一課題研發有效的分析處理方法，以提高地形視覺模擬的效能。本研究使用四分樹的方法建立多層細緻程度(multiple Level of Detail)的地形資料，依使用者視覺條件參數進行判斷，漸變式傳輸資料，即時成像虛擬地形。 使用單一固定門檻值進行LOD的建置，將會造成高程差異劇烈地區區塊過度分割，以及於高差不明顯處分割不足等不合理現象產生。因此本研究提出一套以動態方法決定門檻值的LOD資料產生方法；一方面利用計算不同區塊之最大最小高程，以對數方式形成LOD的門檻；另一方面，並限制每層資料中最大網格面積，以避免產生區域過大而分割不足的現象。資料傳輸方面，則以Client-Server為基本架構，使用連接導向方式(connection -oriented style)進行傳輸。此方法優點在於可確保使用者能正確地獲得有效的資料，且在傳輸發生問題時適時地發出警訊。 本研究測試區為中範圍的台中大坑地區及大範圍的全台灣地區。使用資料為網格形式的數值地形模型(Digital Terrain Model, DTM)，解析度分別為50cm及40m。將地形切割為10×19個區塊(tiles)，每個區塊為513×513 pixels。研究成果顯示，利用本研究研發的動態門檻可有效地解決分割問題，而系統亦可擬真地展示大型的地景視覺模擬。 Cyber city is a system that allows users to inquire about real world information quickly and conveniently on the internet. Terrain is one of the most important factors in the rendering of virtual scenes regarding cyber city visualization. For large-area and detailed representations of digital terrain models, rendering in real-time will be difficult because of the large data volume. It is also difficult to smoothly and efficiently broadcast virtual environments in large-scale web applications. To address these issues, this research developed an efficient technique to progressively transmit and smoothly rebuild cyber city terrain models for internet-based geoinformatic applications. The full terrain was divided into regular tiles to stream adaptively. Different levels of detail (LOD) for each tile were created and stored in the serve end and transmitted to clients according to different visualization conditions. Using static thresholds to build LOD layers may not fully consider different characteristics of terrain tiles, and causes over-sampling or under-sampling. To solve this problem, a dynamic threshold algorithm was proposed to examine the property of each tile to determine appropriate thresholds of the quad-tree process. The first task of the developed progressive transmission was to transmit the base terrain data. Secondly, suitable LOD of each visible tile was determined based on viewing parameters in real time. Change vectors describing differences of adjacent LODs for each tile were then constructed and sent to the client to modify the developed mesh representation of the terrain. Algorithms for the decomposition and reconstruction of subdivision surfaces of tiles were also developed to update each tile as view-dependent parameters changed, so clients could have smooth scene rendering. Experimental results of this study demonstrated that the developed algorithms are effective and efficient in dealing with large terrain visualization. With these techniques, users can reconstruct smooth approximations of the original scenes adaptively starting from a rather small amount of data received on the internet to increase the efficiency of terrain rendering in cyber city visualization.
