Methodologies Used in Identifying Land Cover Types

This paper reviews background, methodology, and accuracy, regarding remote sensing and identifying land cover types. Rational is discussed regarding the use of remotesensing apparatus, multispectral imaging, supervised and unsupervised techniques. The origins of GIS are discussed. Key methods, (supervised and unsupervised), are defined. Comparisons between

digital and visual classifications are made. Spectral mixture analysis and supervised classification geometric correction are discussed. The superiority of visual classification is stated.

A landscape is composed of everchanging elements. Their spatial and temporal patterns distinguish a landscape to an observer; at the same time they inform us of the complexity of dynamic processes at various scales.

For certain remotesensing applications, such as monitoring environmental change, it is essential to be able to compare classes from image to image. In spite of successes in classifying some individual multispectral images using conventional supervised or unsupervised techniques, it has been difficult to obtain consistent classes from images taken at different times, owing to variability in illumination, atmospheric effects and instrumental response. As a result, with a few exceptions (e.g., Hall et al., 1991, and Lucas et al., 1993), monitoring of changes in land cover by remote sensing typically has been restricted to measureme

. . .
hematic feature, such as vegetation type. Surface feature types are verified or selected from ground-based information. On the other hand supervised classification is defined as methods of classification based on external knowledge of the area shown in an image. The supervised methods require input from the user before the chosen algorithm is initiated (Mather 1987). Supervised classification requires the process of using samples of known identity (i.e., pixels already assigned to information classes) to classify pixels of unknown identity (i.e., all the other pixels in the image). The procedure usually is to select training data to create spectral signatures, classify the images involved, and perform an accuracy assessment (Mather, 1987). Exhibit 2 above illustrates the basic differences between supervised and unsupervised classification. Maximum likelihood classification is very sensitive to the number of input variables. Generally, as the number of variables increases, classification accuracy decreases (Peddle, 1993). Kenk et al. reported that classifying ground cover using five or six TM bands instead of four resulted in only marginal increases or in reductions in accuracy. Landgrebe suggested that too many
. . .

Some common words found in the essay are:
Classification Existing, Baca Varela, Types Abstract, Hinton Baker, Settle Wyatt, GIS Kontoes, Anderson Star, Conese Maselli, Correction Geometric, Sensing Vol, remote sensing, et al, land cover, visual interpretation, adams jb, remote sensing vol, classification accuracy, sensing vol, land covers, unsupervised classification, smith mo, photogrammetric engineering remote, journal remote sensing, engineering remote sensing, adams jb smith,
Approximate Word count = 3039
Approximate Pages = 12 (250 words per page)

More Essays on Methodologies Used in Identifying Land Cover Types