圖11 系統刪除誤判崩塌地圖示
(a)編輯修改山崩邊界前 (b)編輯修改山崩邊界後 圖12 系統刪除誤判崩塌地圖示
圖13 設定四種山崩特性之範例示意圖
5. 結論與建議
本研究從半自動化的角度出發，嘗試結合彩色 航照與空載光達資料辨識崩塌地範圍；並且藉由三 維立體展示的人機操作介面，供後續人工判釋及編 輯崩塌地邊界向量資料，以獲取最高品質的成果。
初步結論為：
1. 結合正射航照與空載光達資料，利用彩色航照 的綠度指數及空載光達資料推演的地形坡度、
地物高度模型與地表粗糙度等地形地物指標，
透過人工圈選訓練區自動計算相關指標之門檻；
此種作法隨資料的不同選擇門檻值，可避免不 同地區因不同時間拍攝資料的差異，造成門檻 值決定上的困擾。
2. 自動化及半自動化山崩偵測皆無法達到百分之
百正確率，因此開發了三維立體編修系統，以 對崩塌地邊界向量之端點進行編修，另配合山 崩地質之研究，亦設計了人工判釋及設定山崩 特性之功能，可在判釋所有山崩後，依據山崩 發生之位置、形狀、及崩塌方式等建置崩塌地 的屬性，提供後續之規劃參考用。
3. 測試結果顯示使用自動化偵測崩塌地之成果可 靠度高，且誤判之部分可以透過人工立體編修，
快速判釋是否為崩塌地並刪除之。而評估自動 化辨識崩塌地之困難處，在於崩塌地偵測之邏 輯運算過程雖然簡單，但不同參數門檻值的設 定有許多變因存在；如綠度指數的計算是利用 航照影像之灰度值，會受到太陽光位置、地形 坡度、地物反射特性、大氣狀況、陰影，以及 該地點與攝影機之相對位置等因素影響，所以
無法明確定義出絕對的植被指數。未來在實際
內政部，2005。辦理 LIDAR 測區之高精度及高解 析度數值地形測繪、資料庫建置與應用推廣工
Thackray, and S. J. Dorsch, 2006. “Analysis of LiDAR-Derived Topographic Information for Characterizing and Differentiating Landslide”, Geomorphology, Vol.73, pp.131-148.
Hsiao, K. H., J. K. Liu, C. C. Lau, and J. Y. RAU, 2009, Automation of Landslide Detection Using Optical Images and LiDAR Data, Proceeding of Asian Conference on Remote Sensing, Oct.
18-23, Beijing, China, CD-ROM.
McKean, J., and J. Roering, 2004, ”Objective Landslide Detection and Surface Morphology Mapping Using High-Resolution Airborne Laser Altimetry”, Geomorphology, Vol. 57, pp.331-351.
Parker, J. R., 1997, Algorithms for Image Processing and Computer Vision, Wiley Computer, New York.
1Researcher, Green Energy and Environment Research Laboratories, Industrial Received Date: May. 13, 2010
Technology Research Institute Revised Date: Jun. 21, 2010 2 Assistant Professor, Department of Geomatics, National Cheng Kung University Accepted Date: Jul. 27, 2010 3Senior Researcher, GEL, ITRI
4Chief of Planning Section, Forest Planning Division, Forestry Bureau, COA
*.Corresponding Author, Phone:886-3-5915468, E-mail: [email protected]
Integrating Airborne LiDAR and Color Aerial Photography for Landslide Interpretation
Chi-Chung Lau
1*Kuo-Hsin Hsiao
1Jiann-Yeou Rau
2Jin-King Liu
3Jer-Rong Wu
1Qun-Xiu Huang
4ABSTRACT
In this study, a semi-automatic landslide classification method is proposed and implemented using both airborne LiDAR data and color aerial photographs. A man-machine interface is implemented with three-dimensional perspective display capability, thus to ease the manual interpretation and editing after a preliminary result of landslide polygons are generated by an automatic algorithm proposed in this study. Four parameters are used in the automatic algorithm, namely Greenness, Slope, OHM and Roughness. Greenness is derived from color aerial photograph and employed to define non-vegetated land of fresh scars of shallow-seated landslides. All the other tree parameters are derived from LiDAR DSM and DEM to portray the geomorphometric characteristics of landslides. Thresholds are derived automatically from training areas and then are applied in real time to show the distribution of possible landslides. With the 3D interactive interface, the possible landslides are draped on 3D perspective landscape on the screen. This is convenient to assist the human expert to further modify the results by deleting erroneous ones or modify the boundaries of landslides visually. In addition, the interface system can go without the semi-automatic approach and just be used for visual interpretation solely on basis of the expert knowledge of feature locations, shapes, and types of landslides. Attribute table linked with the landslide spatial feature can also be established in this system. It is proved in this study that this is a practical and efficient system.
Keywords: Landslides, Digital Terrain Model (DTM), LiDAR
Volume 15, No.1, March 2010, pp. 123-140 台數位製圖相機Z/I DMC（Digital Mapping Camera）
及空載多光譜掃描儀 Leica ADS 40（Airborne Digital Scanner 40），有效縮短了後製圖資之流程，
並增廣了影像後端加值應用的效益，一貫且有效工 中勤務總隊所有之Beech-200 及 Beech-350，其中