Chapter 13. Remotely Sensed Multispectral Scene Analysis

Summary

The classification of remotely sensed data involves the procedures for assigning the multiband image pixels obtained from remotely sensed satellite imageries to an appropriate set of meaningful object classes. From the classified images, useful information can be made available for the proper utilization of natural resources. In this chapter we have presented the descriptions of various satellite sensors and satellite imageries. The spectral reflectance of various categories of earth objects, such as soil, water body, vegetation and so on have been discussed here. Neural network based approaches towards remotely sensed scene classification, such as classification using multilayered perceptrons with error back propagation and counter-propagation networks have been adequately discussed here. The design of a knowledge based approach for spectral classification of images has been elaborated here. Spatial reasoning plays a useful role in interpreting remotely sensed images. The generation of a set of spatial rules using the domain knowledge of the scene has also been discussed here. Finally applications of fuzzy set theory in remote sensing has been discussed briefly.