台灣地區近年來陸續發生多起因工業廢棄物、廢水、或其他污染源所造成的重金屬污染事件。過去對於這些污染事件的查證與評估多仰賴人工現場勘查,不僅需耗費大量的人力與資源也無法迅速有效地達成令人滿意的成果。利用遙測資料進行重金屬污染農地的調查雖然較有效率,然而傳統遙測資料由於僅具備少數波段的資訊因此必須配合大量的輔助資料,也只能透過污染源對場址上各種地物所造成的影響進行非直接性的分析與判釋。 近年來日漸普及的高光譜遙測影像不僅可提供豐富的高解析力光譜資訊,其具備的近乎連續性光譜特性更提供了進行比傳統遙測研究與應用更精密、更複雜分析與應用的可行性。本研究即研議利用高波譜解析度衛星遙測影像進行重金屬污染農地辨識之應用。透過高波譜解析度遙測影像所提供的豐富波譜資料,針對不同重金屬污染源的波譜特性進行深入的分析以迅速且有效地自影像中判識出受重金屬污染的農地區域。研究的重點將置於開發一可充分利用高波譜解析度影像特性與優點的影像分析法則與流程,以建立一套更經濟、更有效率的農地重金屬污染分析與評估方式。 Heavy metal contamination in agricultural lands has become a serious threat to the health of citizens, economics development, environmental protection, and ecological conservation works in Taiwan. Until relatively recently, the authorities have to depend primarily on in situ mechanisms to investigate heavy metal contamination incidents, which often require a lot of manpower and resources and are not easy to reach reliable results timely. Although remote sensing has become more and more popular and has played a more important role in environmental disaster investigation and evaluation, traditional sing-band and multispectral remote sensing data can only be used for indirect analysis because of the limited information they can provide in a few discrete spectral bands, if applied to investigations of agricultural lands contaminated with heavy metals. On the other hand, images acquired with hyperspectral sensors not only can provide data in tens to hundreds of spectral bands, the spectral details and features produced from them also allow researchers and investigators opportunities to perform sophisticated analysis and applications that are difficult to achieve using traditional multispectral remote sensing data and techniques. Combining with other remote sensing and reference data, hyperspectral imagery should be more suitable and more productive in image analysis for identifying heavy metal contaminated agricultural lands. This research targets on developing an algorithm and procedure of hyperspectral image analysis in order to detect heavy metal contaminations in agricultural lands. The algorithm and procedure developed in this study will take advantage of the characteristics of hyperspectral data, and will effectively and efficiently identify areas contaminated with heavy metals from satellite hyperspectral remote sensing images. Therefore, they will provide a more reliable and cost-effective approach for the investigation, evaluation, and monitoring of heavy metal contaminations in agricultural lands. 研究期間:9207 ~ 9307