ABSTRACT
Remote sensing imagery needs to be converted into tangible information that can be used together with other datasets in geoprocessing. The objective of this chapter is to obtain remote sensing data and perform unsupervised segmentation for pattern recognition of multispectral imagery. Pixel patterns are detected by segmentation and propose a classification scheme for landscape targets from normalized difference vegetation index. A segmentation approach is applied to RGB bands to simplify regions with similar targets and recognize R, G, B values in these regions. RGB indices can be determined for each super-pixel to form a geographic database of the region. Solved exercises are used to illustrate concepts about unsupervised classification and image segmentation. Calculations for determining Euclidean distance from simulated pixel values are presented as a simple approach to understanding the clustering process.
