Classification of Agricultural Crops in Radar Images

For the past few years an accurate X-band SLAR system with digital recording has been available in The Netherlands. The images of this system are corrected to indicate radar backscatter coefficients (gamma) instead of arbitrary greytones. In 1980 a radar measurement campaign was organized in the Flevopolder area. Ground-based measurements were carried out and at certain prefixed times flights were made. The optimum flight times had been derived from a simulation study, based on previous measurements. The results of this campaign are now being analyzed. The one-dimensional case, consisting of one flight over the test area at a previously derived optimum flight time gives only 35-percent correct classification in 7 classes. Multidimensional cases with flights at different dates improve the classification result to over 85-percent. The small differences in backscatter between the various types of crops clearly demonstrate the need for an accurate radar system which does not show too much speckle. The results of these analyses are reported in the following.