Land-cover classification using SIR-C/X-SAR data

With the flight of the Shuttle Radar Lab (SRL-1) in April and of SRL-2 in October of 1994, polarimetric SAR data is now available from a spaceborne sensor. While the previous efforts have been impressive, this new sensor promises even more discrimination ability. Once this sensor is made into a free-flyer (satellite) the algorithms developed now should work with minor modifications, allowing the data from the satellite to be used almost as soon as it is put into orbit. Consequently, it is important to devise a classification algorithm that is robust to variability in seasonal variations, weather, and local vegetation species. This paper presents a first step toward that goal. Several scenes from the April, 1994 flight are to be classified using a knowledge-based, hierarchical classifier. During the flight the snow cover changed remarkably as did the amount and distribution of standing water over the Raco, Michigan site. Despite this variability, a single classifier is to be developed that can classify five or six structural classes (bare surfaces, short vegetation, and a few different kinds of trees) with high accuracy. This has been achieved with a single image with accuracies higher than 90%. Other data from the October, 1994 flight are also to be classified and the differences between the two classifiers is to be related to changes in physical parameters between the two seasons. For example, since the deciduous trees do not have leaves during the winter, the C- and X-band radar responses are quite different in the summer than in the winter. Basic knowledge such as this can be used to adapt the classifier to seasonal changes