利用高光譜影像偵測外來植物

Hyperspectral Image Analysis for Mapping Invasive Plant Species

蔡富安(Fu-An Tsai) ； 林恩楷(En-Kai Lin)

《科儀新知》 161期 (2007/12) Pp. 5-11

https://doi.org/10.29662/IT.200712.0001

摘要

隨著遙測技術的進步，遙測影像的光譜頻道從以往單波段及個位數波段的多光譜提升至數十至數百個波段的高光譜。高光譜資料提供了大量、高光譜解析力且近乎連續性的光譜資訊，因此有極大的潛力進行如特定物種分類等複雜的分析及應用。然而高光譜影像的資料量和其高維度的特性，使得使用者可能無法有效地使用一般傳統的（多光譜）影像處理方法進行分析。主成分分析(PCA)為影像分析時增顯資料差異，以萃取特徵並降低資料維度的常用方法之一，但並不一定適合直接應用於高光譜影像，因其在特徵萃取時有可能忽略差異較小但卻有助分類的資訊。因此本研究針對植物分類發展一光譜區段式PCA，以更有效區隔不同植物間之差異，正確分類判釋外來植物銀合歡在南台灣恆春地區的覆蓋範圍。

並列摘要

Hyperspectral remote sensing images provide more complete and detailed spectral information about ground coverage and have a great potential for more sophisticated applications. However, the large data volume and high dimensionality of hyperspectral data can cause substantial impact to its applications. Principal component analysis is a common technique used for feature extraction and reduction in remote sensing image analysis, but it may overlook subtle but useful information when applied to hyperspectral images. This research developed a spectrally segmented principal component analysis scheme that can not only reduce the dimensionality of a hyperspectral image but also extract critical spectral features helpful in discriminating different vegetation types. The developed methodology was applied to the analysis of an EO-1 Hyperion hyperspectral image to determine the status of an invasive plant species (Leucaena leucocephala) in southern Taiwan.