應用光譜水體指數結合數學形態學於光學影像進行河道變遷偵測之研究

臺灣本島為一南北狹長、東西窄的島嶼，造就河川有流路短促、坡陡流急的特性，夏季熱帶氣旋對流旺盛，豪雨沖蝕使得中下游河道易發生沖淤的現象；加上全球氣候極端化，河川洪枯流量越趨懸殊，河川深槽處於不穩定的狀態，使得河道變遷更加頻繁與複雜。對於河川規劃治理與災害防治，首要基礎工作則是河道變遷分析。欲瞭解河道變遷的情形，過去多以河道大斷面測量為之，近代則發展以遙感探測技術達成河道變遷偵測，如航空影像或衛星影像等，隨著新興技術的發展如無人機(Unmanned Aerial Vehicle, UAV)或光達(Light Detection And Ranging, LiDAR)亦可運用於河道變遷偵測。上述各項技術各有其限制，由於光學影像如Landsat和Sentinel具有免費下載、定期定點拍攝之優勢，故本研究選擇光學影像作為研究資料。本研究以臺灣本島北、中、南、東各區河川，分別為淡水河、北港溪、曾文溪、旗山溪與秀姑巒溪作為研究區域，計算光譜水體指數(Spectral Water Index)並加入數學形態學(Mathematical Morphology)的概念，藉其能快速並準確提取水體之特性，同時完整萃取河道的邊界，進一步分析河道的變化。研究成果顯示：Normalized Difference Water Index(NDWI)與Modified Normalized Difference Water Index(MNDWI)適用於旗山溪、秀姑巒溪與淡水河其河川特性為辮狀型態之河川，Automated Water Extraction Index(AWEI)則適用於北港溪與曾文溪其河川特性為蜿蜒型態之河川，由此可知根據不同河川型態特性，各種光譜水體指數的應用仍具有差異，因此本研究之貢獻主要為針對臺灣本島河川，歸納各種光譜水體指數於地表水體提取之適用性。

關鍵字

河道變遷偵測；光學影像；光譜水體指數；數學形態學

並列摘要

The characteristics of rivers in Taiwan are the steep slope with high sediment concentration. The distribution of precipitation is non-uniform due to the geographic environment and extreme events. Moreover, with the condition of global climate change, the dynamics of channel meandering become complicate and frequent. For river governance and disaster prevention, the primary work is the analysis of river change. To achieve river change detection, field measurements and remote sensing technology are necessary. With the development of new technology such as UAV and LiDAR, they can also be used for river change detection. Because optical images have the advantages of revisit time and long-term data collection, this study takes optical images as dataset. In this study, the study area includes Tamsui River, Beigang River, Zengwen River, Qishan River and Xiuguluan River. This study combines spectral water index and mathematical morphology to capture water bodies based on multitemporal optical images. In addition, this study delineates the river channel to analyze the change of river. The results show that Normalized Difference Water Index(NDWI) and Modified Normalized Difference Water Index(MNDWI) are suitable for braided rivers such as Qishan River, Xiuguluan River and Tamsui River. Automated Water Extraction Index(AWEI) is ideal for meandering rivers such as Beigang River and Zengwen River. As a result, this study summarizes the applicability of each spectral water index for surface water extraction according to various river types in Taiwan.