合成孔徑雷達影像於颱風豪雨後淹水之偵測

合成孔徑雷達影像（SAR）可以穿透雲霧、日夜皆可運作，可克服光學遙測影像在不良天候觀測的不足，對於颱風豪雨後的災情偵測有相當的優勢。本研究先以SAR影像對曾文水庫的水體進行辨識，進一步配合理論雷達陰影區與相對高程模型HAND（Height Above the Nearest Drainage）之資訊，可大幅降低誤判。結果顯示水體的辨識精確性（F1-Measure）從66.6%提升至92.0%。淹水區域則可透過兩張SAR影像中相對應像素間的差異作為判定，這個差異在6.2 dB有最佳的辨釋精確性約82.6%。本研究提出一個SAR影像淹水偵測流程，並以實際水災事件作為案例，進行淹水範圍影響之估測，分析的結果對於應變災情資訊提供有相當助益。

關鍵字

遙測；合成孔徑雷達；淹水

並列摘要

Comparing to optical remote sensor, Synthetic Aperture Radar (SAR) is an active sensor and radar signal can penetrate clouds for working in near all-weather/day-night. Therefore, SAR damage assessment methods enable useful disaster response in typhoon and heavy rainfall. This study first uses SAR images to detect change of surface water area in Zengwen Reservoir and limits surface water areas by HAND (Height Above the Nearest Drainage) mask and radar layover & shadow mask. The result's F1-Measure increase from 66.6% to 92.0% after using mask data. Backscatter decreases due to totally flooding, therefore threshold values can be used for separating flooded area. The best detected result: F1-Measure 82.6% occurs when the threshold is 6.2 dB. This study proposes a flood detection process by SAR image, and uses actual flood events as a case to estimate the impact of flooding range. The results of the analysis are quite helpful for providing the assessment and quick response after a hazardous event.

並列關鍵字

Remote Sensing ； SAR ； Inundation

國際替代計量

合成孔徑雷達影像於颱風豪雨後淹水之偵測

全文下載

主題瀏覽