An overview is presented of applying the Neural Network-Self Organizing Feature Map (SOFM) to the problem of classifying hyperspectral remotely sensed data. Remote Sensing satellite sensor systems that passively image the earth have been used to gather geospatial data of the earth since the 1970s. The systems that currently orbit the earth collect information in three to seven segments or bands of the electromagnetic spectrum. These systems are called multispectral systems. Airborne sensor systems are now being developed that collect data in many more bands of the electromagnetic spectrum. An example of these types of sensor systems is the AVIRIS sensor flown by JPL. It collects 224 spectral bands of information. These systems are called hyperspectral sensor systems. Neural Networks simulate biological interaction of neurons in hardware and software to develop learning procedures that perform analysis functions. The Self Organinzing Feature Map (SOFM) is a type of Neural Network that uses competitive learning to cluster and organize data.

WHERE: TEC 340