由於氣候變遷及極端降雨之影響，近年台灣坡地災害頻繁。為了山崩災害之相關研究及地質災害敏感區劃設之需求，須使用合適之斜坡單元，建立各類型山崩潛勢評估系統。本研究先擇定大漢溪流域內之一小集水區做為訓練區域，以物件導向式分析方法自動化產製斜坡單元。利用數值地形模型計算研究區域內之坡度、坡向、曲率、水系、集水區等網格資料圖層，匯入物件導向分析軟體 Definiens中，給定各資料圖層權重後，根據影像資料之光譜均調性及空間相關性進行影像分割，產製合適之斜坡單元。隨後以此影像分類準則應用於大漢溪流域其他小集水區進行檢核，檢討其成效後，建立合適之影像分類通則，並進行各斜坡單元內坡向標準差之統計，以做為評定成果優劣之用。與使用其他因子之分割程序相比較後，最後使用坡向圖層輔以集水區及水系圖層進行影像分割，使其滿足各斜坡單元之內部差異須最小且單元間差異須最大之基本定義。分析結果顯示，一級河之集水區大多可區分為二至三個或更多個斜坡單元，與前人研究結果相比，能獲得更細緻而合理之劃分結果。 Slope disasters have received attention in recent year worldwide and in Taiwan, because of extremely heavy rainfall presumed to relate to climate change. For the purpose of landslide research and geological sensitive zone designation, we require division of appropriate slope units for building up various types of landslide potential evaluation system. In this study, one of a small catchment in the Tahan river watershed was chosen as the training area, using an object-oriented classification method to automatically generate slope-units. I used DTM (Digital Terrain Model) as raw data to produce raster data layers which include slope gradient, slope aspect, curvature, drainage system, catchment area. These raster data were imported into an object-oriented analysis software, which was known as Definiens, weight of each data layer was given, and then segmentation as well as slope-unit generation was executed, according to the heterogeneity and spatial correlation of the imported data. After that we applied this rule in other catchments of the Tahan river watershed and viewed its effectiveness to make further amendments for a reasonable general rule. A standard deviation of slope aspects in each slope-unit was calculated for checking purpose. Finally, we used the slope aspect and the catchment area, as well as drainage system for segmentation through comparison of the results from different layer weightings. The final result of using this segmentation criteria satisfied the definition of slope-unit that high homogeneity within each unit and high heterogeneity between the neighbor units. It shows that each of a small catchment of a first-order stream could be divided into two or three or even more parts of slope units. This provides more detailed and reasonable result than previous studies.
