利用雷達強度變遷偵測邊坡變動

臺灣位處板塊交界帶，地質狀況破碎且地形起伏大，故山區邊坡常有變動發生，又屬山崩(Landslide)為最嚴重的災害。在臺灣山崩常由地震、強降雨所引發。近年受到極端氣候影響，山崩發生的次數、強度提升。因此，對於山崩災害的長期追蹤是刻不容緩的議題。由於雷達影像具有不受天氣、雲霧限制的優勢，能穩定、定期獲取變動資訊，且其強度資訊能克服InSAR技術在山區同調性低的限制。故雷達強度變遷偵測為本研究主要使用的方法。與前人研究不同，本研究會在斜距(Slant range)下進行強度變遷偵測以降低影像的幾何扭曲，並以物件式影像分析(Object-Based Image Analysis，OBIA) 系統地萃取有山崩有意義的遙測識別標誌(Remote sensing signatures)。此山崩偵測流程能提供概略的山崩範圍、位置，且成果顯示被萃取的山崩特徵與山崩目錄吻合良好，尤其是大面積的崩塌地。另外，本研究基於高時間解析度的Sentinel-1影像制定一山崩長期監測策略，它能提升山崩偵測的能力，並提供山崩在時間、空間上的訊息。本研究亦透過時間序列的遙測識別標誌發現新生崩塌地。總結來說，本研究建立了以雷達強度影像為基礎的山崩偵測流程，並提出整合多期遙測識別標誌的長期監測策略。它們增加了雷達影像應用於山崩偵測的可行性與使用強度變遷偵測方法的新方向，並以長期監測的方式提供邊坡變動的預警。

關鍵字

山崩；衛載雷達影像；遙測識別標誌；雷達強度變遷偵測；物件式影像分析

並列摘要

Landslide is one of the common slope change types occurred in Taiwan and is easily triggered by typhoons or earthquakes. To secure life and property, long-term monitoring of landslide disaster becomes a critical task. To achieve this, SAR intensity change detection was introduced in this dissertation. to overcome potential weather limitations using optical images and possible low coherence in InSAR processing. Different from others, we conducted SAR intensity change detection in slant range image for mitigating geometric distortion. Object-Based Image Analysis (OBIA) was applied in maps of SAR intensity change to extract potential regions of landslide. With rapid processing, rough indicators for providing coverage and location of landslide were derived. The results show that SAR remote sensing signatures of landslide match well with landslide inventory. Based on the verified method, a long-term monitoring strategy was further developed using Sentinel-1 images because of its short revisiting time and continuous updating. It was found that the time-series results enhanced the capability of landslide detection and provided spatial and temporal information of landslide. We achieved a systematic landslide feature extraction using SAR signatures and successfully developed the strategy of integrating time-series landslide detection results. The novel scheme increased the feasibility of utilizing SAR intensity information on landslide detection and is of potential to provide early warning of slope change.