Laserfiche WebLink
<br />000054 <br /> <br />Classification <br /> <br />Digital image analysis techniques have several advantages compared <br />with visual techniques. They are generally faster, provide objective <br />decisions, and capitalize on the multiple bands and high radiometric <br />resolution of TM data (Meyers, 1983). The digital nature of the source <br />image data is also directly compatible with GIS databases. <br /> <br />The overall objective of image classification procedures is to <br />automatically categorize all pixels in an image into land cover classes or <br />themes. . This is accomplished by identifying different feature types based <br />on their inherent spectral reflectance or emittance properties (Ullisand and <br />Kiefer. 1987). One type of spectrally oriented classification procedure is <br />the unsupervised classification technique. . Unsupervised classifiers involve <br />algorithms that examine the uJjknown pixels man inlage and aggregate <br />them Into a ;J.lQ.P1ber of~lassesb~sed on the.natural groupings or clusters <br />presentintlleilnageVa1tles(Ijllisand and Kiefer, 198.7). 'Clustering allows <br />. the analyst to qUlcldy-identifymany. cla$ses and to detect classes that are <br />not in contiguous, easily reco~zed regions (ERDAS, 1991). <br /> <br />:All image processing for the Upper 'Colorado Irrigated Lands Project is <br />being:accomplishedwithERDAS software. The particular algorithm used <br />for thi'Sstudy is aniterative technique referred to as ISODATA(Iterative <br />Self-Organizing Data Analy~is Technique). The ISODATA process begins <br />with a specified.nul11ber of arbitrary cluster me~sand process'es them <br />repetitively. tQ . shift thearbittaty means. to' the means of the clusters in the <br />data (ERDAS, 1991). As an iterative technique, it is not biased toward the <br />top of the file, as in one-pass algorithms. <br /> <br /> <br />Upon completion of the clustering program, a set of spectral <br />signature files is produced. Eachsigliature is a setof statistical data that <br />defines an individualspectl'al. cluster. These signatures must be evaluated <br />to determine the validity of the clustering results. One evaluation <br />technique plots tile normaldjstributions for each two band combination in <br />two-dimensional spectral space. When plotted in this way, each potential <br />class results in an ellipse. Wh.en the ellipses have extensIve overlap, the <br /> <br /> <br />...i(;~.....~:i'F~,~~tlll:rlli~.~#I& <br /> <br /> <br />uliclassified~ The iinage analyst Itiustdeterfulliethe maximuni' number of <br />classes the data can support and to relate those spectral classes to land <br />cover types. Parameters such as the number, size, and separation of <br />clusters can be specified by the analyst. <br /> <br />8 <br />