METHOD FOR CLASSIFICATION OF UNDERLYING SURFACE ON BASIS OF MULTICHANNEL IMAGE Russian patent published in 2008 - IPC G06K9/62 

FIELD: methods for digital processing of images, possible use for classification of underlying surface in multi-channel systems for remote probing.

SUBSTANCE: method includes forming an image of underlying surface in q spectral channels with various location sensors, performing geometrical correction and presentation of component images in unified system of spatial coordinates, density of brightness value probability is transformed to gauss form of those component images in which it is different from gauss form, class number of underlying surface is determined for each element of multi-channel image, and when generating component images of underlying surface for every element of component image, a posteriori probabilities of association of given element of component image with k class of underlying surface (k=1,2,...,K) are determined, for every element of multi channel image a posteriori probabilities of association of that element with k class of underlying surface are determined by addition of corresponding a posteriori probabilities of component image and their normalization to number of channels q, class number of underlying surface is determined for each element of multi-channel image on basis of maximum value of a posteriori probability of association of given element of multi-channel image with k class of underlying surface.

EFFECT: increased precision of underlying surface classification on basis of multi-channel image.

1 dwg