Using a Hybrid FNN Method for Image Classification of Satellite Remote Sensing Data

Satellite remote sensing images play important roles in many practical applications, including meteorology, natural resource identification, ecology, agriculture, emergency and disaster management, as well as mapping and surveying. With the rapid development of the aerospace industry, more and more onboard imaging systems are being constructed and launched, many with hyperspectral and high-resolution capabilities. To fully use the huge amounts of remotely sensed data provided by these systems requires appropriate algorithms, developing these is an ongoing challenge for researchers in academia and in industry. In this chapter the basic concepts of satellite remote sensing image analysis are discussed and the fuzzy neural network (FNN) approach described, both in the context of hyperspectral/multi-spectral images and in the context of images based on the red, green, blue colours of visible light. Experiments based on real-world examples of such images are carried out to illustrate the methods involved. The results are analyzed and show that the FNN approach can give good results for both multi-spectral and visible light images.