Multiresolution Modeling and Terrain Visualization

Large-area terrain visualization is important to the applications of environment simulation and geographic information system. A general personal computer cannot process and render such a large amount terrain data in real time. Thus the most challenge problem to this study is to render and demonstrate such a large terrain model in a reasonable frame rate. We have divided the study of terrain visualization and its applications into seven levels: (i) triangulation, (ii) multiresolution modeling, (iii) view-dependent rendering, (iv) refinement of multiresolution modeling, (v) dynamic loading, (vi) distributed processing and visualization, and (vii) The management and applications of multiresolution modeling. In these years, we have spent much efforts on the studies of triangulation, multiresolution modeling, view-dependent browsing, and dynamic loading.

Currently, we spend our efforts on the refinement and applications of multiresolution terrain modeling, such as appearance preserving multiresolution modeling, city scope visualization, flight simulation, replay the flight behavior and path, ground surveillance and tracking on the air. We also study how to provide the immersive control equipments such that we can watch the stereo view of the environment and control the objects using 3D tracker and sensing glove.

The characteristics of our studying are that all programs and applications software are written and developed by the members of this lab. and we only use  Microsoft visual C++ languages, Microsoft MFC basic library, and OpenGL graphic library. We don't use the commercial software or development toolkits. The development procedure is hard, but we keep the key techniques such that we can match various requirements on applications.

For these studying topics, we must have the related geographic data, and the most important data are the digital terrain model and satellite images. We have the digital elevation model of the whole Taiwan; it is 200 mega-byte binary file. We provide a clipping program to clip any necessary area of the model. The clipped terrain model can be converted and safe as a nff, dat, mff, or  mbf model file. We also have the SPOT Pan/XS 6.25 m synthesized color images of the whole Taiwan.  We also provide a clipping program to clip any necessary area of the satellite image. The clipped satellite image can be converted and safe as a jpg or bmp image file.

 

Results: Terrain visualization, Multiresolution modeling, View-dependent rendering,

Dynamic loading, Tactical simulation

 

Members: 李岳倫林君威、楊璿、林明杰、張進豪、黃威凱、劉世昌、

洪寶淑、楊天行、蔡淑齡、譚國俊、黃崇杰、石家豪、劉文鑫、

黃書政

 

Terrain visualization

下圖是一塊 25.6 ´ 25.6 平方公里的台東地形。如果沒有虛擬實境技術的支援,則展示這一塊地的速度 (在一般的個人電腦上) 大約是每秒 0.12 個畫面 (也就是 8.3 秒才能展示一個畫面)。而在虛擬實境技術的支援下,高程誤差控制在 5 %  以內,我們可以以每秒 24 個畫面的速度來展示這一個地形。也就是說,展示速度提高了 200 倍。

 

 

Multiresolution modeing

多重解析度模塑的意義是先將地形模型塑造成一些離散 (discrete) 或連續 (continuous) 變化解析度的模型。當瀏覽該地形時,展示系統即配合觀看的距離或誤差來選擇適當解析度的模型做展示。例如飛行模擬中,當飛機飛得高看得廣,系統即選擇或即時變化較低解析度模型來展示;當飛機飛得低看得窄時,系統即選擇或即時變化高解析度模型來展示;如此即可提高瀏覽速度又不會降低觀看的品質。例如,下圖石門水庫模型

 

32,768 triangles.                                                        8,109 triangles.

 

1,978 triangles.                                                              467 triangles.

 

View-dependent browsing

視點相關展示的意義是當我們在瀏覽地形時,系統會根據視點參數 (view parameters) 來自動調整模型中各區域的解析度做展示。例如,看不到或距離遠的區域,系統會將該區域的模型變粗糙;當粗糙區域進入視線中或距離拉近後,系統又會逐漸提高該區域的解析度。因此相同複雜度的模型在具有視點相關功能的系統中可展示更高的幾何精密度或 得到更快的展示速度;例如,下圖翡翠水庫模型 (19,718 triangles) 。

 

  

視點無關                 視點相關

 

 

Dynamic loading

動態地形區塊載入的意義是說當地形模型大到無法放到主記憶,而需先將模型分割成一小塊一小塊的,然後以多重解析度模型模塑的技術來簡化各個小區塊模型,並且將簡化的過程記錄下來。當需要用到那些區塊時,則依據目前瀏覽視點的位置及方向決定那些區塊應該從週邊儲存體中載入或從記憶體中移除;並且在視點移動之同時,決定每區塊所應展現的解析度。例如下圖顯示載入翡翠水庫地區的 9 塊地形模型,配合多重解析度及視點相關等技術做即時動態地形展示。

 

 

 

Tactical simulation

這是一個以多重解析度地形模塑技術配合電腦繪圖及目標追蹤等技術來完成的一個地面飛彈基地防空模擬遊戲。我們以某區域 (12.8 km´12.8 km) 為展示區。在其山坡地建立一個飛彈防空基地,並構建會自動旋轉的雷達站、會自主飛行的飛機、會追蹤飛機的飛彈、及可操控方向的觀測窗等。