Learning class descriptions from a data base of spectral reflectance with multiple view angles

A program was developed to learn class descriptions from positive and negative training examples of spectral, directional reflectance data taken from natural surfaces such as bare soil, natural vegetation, or agricultural vegetation. The learning program combined a form of learning referred to as a learning by example with the generate and test paradigm to provide a robust learning environment that could handle error prone data. The learning program was tested by having it learn class descriptions of various categories of percent ground cover and plant height. These class descriptions were used to classify an array of targets. The class descriptions in this program comprised a series of different relationships between combinations of directional view angles, e.g., (30,50), (45,60), (10,180), etc. Where the values in parentheses are for zenith and relative azimuth view angles for a particular view. The program found the sequence of relationships that contained the most important information that distinguished the classes. The concept being learned was a sequence of relationships that optimized the discrimination of a class