Identification of cloudiness by nonparametric pattern recognition algorithm from AVHRR data

Abstract

Within the framework of resource-ecological monitoring of the Earth's surface, the problem of recognition of cloud types from NOAA AVHRR data is considered. Based on the statistical approach, a nonparametric algorithm is developed for recognition of stochastic fields, as well as the procedure for estimation of informative indicators from the condition of minimum estimated recognition error. The conditional probability models of recognized images are reconstructed from learning samples selected by an experienced operator. Classification of the types of cloud fields is exemplified by the NOAA satellite data.