運用福衛二號高時空分辨率多光譜影像於台灣全島崩塌地判釋與災害分析

Using the FORMOSAT-2 High Spatial and Temporal Resolution Multispectral Image for Analysis and Interpretation Landslide Disasters in Taiwan

林恩如(En-Ju Lin) ； 劉正千(Cheng-Chien Liu) ； 張智華(Chih-Hua Chang) ； 鄭依凡(I-Fan Cheng) ； 柯明勳(Ming-Hsun Ko)

《航測及遙測學刊》 17卷1期 (2013/03) Pp. 31-51

https://doi.org/10.6574/JPRS.2013.17(1).3

摘要

台灣每年同時受到三項以上自然災害威脅的人口面積比例高居世界第一位，其中主要為山坡地災害，且台灣缺乏長時間的崩塌目錄。故本研究將2005至2012年間的福衛影像經錯位修正、正射糾正、色彩融合、邊緣銳化與可適化對比增揚處理後，再以「遙測影像崩塌及陰影區專家輔助圈繪系統」進行陰影與崩塌區之判釋、以最小集水區的觀點群組崩塌地、結合陰影資訊的變異分析法找出年間變動量，最後建置崩塌目錄。在崩塌判釋結果方面，2m解析度判釋結果與25cm航照判釋結果相比，崩塌判釋之平均整體準確度達98%。

關鍵字

衛星影像處理 ； 崩塌地判釋 ； 陰影判釋 ； 年間變動量 ； 崩塌地群組

並列摘要

Taiwan ranks top one of the world in the ratios of population and area that are threatened by more than three natural disasters. Among various types of disasters, the landslide occurred in mountainous areas is the major one. To gain an insight of the mechanism and the long-term trend of landslide, we need a detailed landslide inventory of Taiwan generated from the same source of data, using the same interpreting method, with a comparable precision. Formosat-2 images of Taiwan area taken from 2005 to 2012 are processed by Formosat-2 automatic image processing system (F-2 AIPS). This system is able to digest the raw data of Gerald format, apply the basic radiometric and geometric correction, output the level-1A product, conduct the rigorous band-to-band coregistration, automatic orthorectification, multi-temporal image geometrical registration, multi-temporal image radiometric normalization, spectral summation intensity modulation pan-sharpening, edge enhancement and adaptive contrast enhancement, and the absolute radiometric calibration. The expert landslide and shaded area delineation system (ELSADS) is then used to produce the shadow and landslide inventory. Based on the concept of minimum catchment, all landslide polygons are grouped to investigate their annual variations. The accuracy of landslide inventory is carefully validated with the manual interpretations from aerial images (25cm), and the mean accuracy is 98%.

並列關鍵字

Satellite image processing ； interpretation of landslide ； interpretation of shadow ； variation ； group of landslide