Detection and identification of objects on multispectral satellite images

This article considers the detection of objects on the time sequences of satellite multizone images. Optimal and quasi-optimal detection algorithms are built on the basis of a combination of non-linear double stochastic filters and pseudo-gradient procedures. The analysis of behavior of the synthesized algorithms at processing of the real satellite material under conditions of a priori uncertainty concerning parameters of deformation of the reference image is carried out. (paper)