Content area
Full Text
ABSTRACT: The availability of hyperspectral images expands the capability of using image classification to study detailed characteristics of objects, but at a cost of having to deal with huge data sets. This work studies the use of the principal component analysis as a preprocessing technique for the classification of hyperspectral images. Two hyperspectral data sets, HYDICE and AVIRIS, were used for the study. A brief presentation of the principal component analysis approach is followed by an examination of the information contents of the principal component image bands, which revealed that only the first few bands contain significant information. The use of the first few principal component images can yield about 70 percent correct classification rate. This study suggests the benefit and efficiency of using the principal component analysis technique as a preprocessing step for the classification of hyperspectral images.
KEYWORDS: Hyperspectral images, image classification, land use, principal component analysis
Introduction
The spectral resolution of a sensor determines much of the capability and performance of a remote sensing system, which uses the detected spectral properties of the object for processing and analysis. The spectral resolution refers to the spectral width that a sensor can detect in one single image band. Several types of images with different spectral resolutions have been identified (Schowengerdt 1997). The common panchromatic image records the object in one band which covers the entire visible portion (ca. 300 nm wide) of the spectrum. It is therefore known as broadband image. Multispectral images, e.g., SPOT and Landsat images, have a relatively narrow spectral width of about 50-100 nm. Hyperspectral images are those having 5-10 nm spectral width, which can reach a nearly contiguous spectral record for the object. Though future development will lead to ultraspectral images at a spectral resolution of less than 5 nm, recent research activities are mostly focused on multispectral and leading increasingly toward hyperspectral images. Currently available airborne hyperspectral imaging systems are, among others, AVIRIS (Airborne Visible/Infrared Imaging Spectrometer (Porter and Enmark 1987)), HYDICE (Hyperspectral Digital Image Collection Experiment), DAIS (Digital Airborne Imaging Spectrometer (Lanzl and Mueller 1999)), HyMap (http://www.intspec.com, Australia), MAIS (Modular Airborne Imaging Spectrometer), and Push-broom Hyperspectral Imager (PHI) (Zhang et al. 2000). By the time of writing this article (January 2002), the space-borne hyperspectral sensor...