- 學位論文
以遙測對崩塌地影像判釋之研究:螞蟻聚類演算法為實證
The Study of Remote Sensing on Landslides Image Classification through Adaptive ant Clustering
摘要
台灣地區因為地質、地形、氣候等影響容易導致沖蝕、土石流、崩塌等坡地災害，並因為台灣地形的因素，水資源不易保存，水庫是我們重要的飲水來源，不過過多的崩塌地，會導致水質濁度增加與水庫淤積以致水庫的使用年限大幅的降低，故如何快速的監測水庫周圍崩塌地乃一門重要的研究課題。本研究的主旨在以螞蟻聚類演算法方法進行分析崩塌地的影像分析其目的在改善聚類優化的問題。此方法的優點在運算的過程簡單且為非監督式方法不需資料類別，其步驟詳細的說可分為(a)初使螞蟻個數 (b)計算螞蟻選擇群聚 (c)更新費洛蒙 (d)計算新群心(e)進行迭代，以此方法計算萬大水庫崩塌地得到的結果為正確率為82% 且崩塌的主題圖已繪製， 可準確研判崩塌的的位置，且非監督式地計算成低，故此法有助於以衛星影像監督崩塌地的影像分類問題的判釋。
並列摘要
Geology, topography and climate can easily induce debris-flow, erosion and landslide. On the other hand, the terrain is complex that the water resource can be reserved very difficultly. The massive landslide will produce the turbid of the water quality increase and the deposit will become worse. Hence, the lifetime usage of the reservoir will be reduced. To develop a monitoring system to investigate the landslide surrounding reservoir is a crucial work. In this study, the ant-clustering algorithm is developed. The advantage of this algorithm is to effective cluster data into groups. The steps are (a)initialize the number of ants (b)compute the ants to each cluster center (c)renew the pheromone (d)recalculate the cluster center (e) do iterations. The accuracy rate of this method is 82% of landslide classification. The landslide thematic map is drawn and the position of occurrence place is shown. The advantage of this process is low cost and it is very effective comparing to other supervised learning approaches.