METHOD OF RECOGNISING IMAGES OF COMBINED NATURAL AND MAN-MADE ENTITIES AND ESTIMATION OF STATUS PARAMETERS THEREOF FROM AEROSPACE PROBING HYPERSPECTRAL DATA Russian patent published in 2011 - IPC G01V9/00 

FIELD: physics.

SUBSTANCE: aerospace probing hyperspectral digital data are received and recorded onto a magnetic carrier. Obtained data are processed via geographical referencing. Geometric, radiometric and spectral distortion of data is taken into account. A specific spectral channel with maximum distinctiveness of classes is searched for based on the bar chart of brightness of the viewed objects. The maximum number of distinguishable classes in that spectral channel is found. Resolution elements corresponding to the average (reference) spectrum are identified. The elements are identified for each selected class and the introduced information measure of variability of that spectrum. Possible channel correlation is considered during identification. An automated classifier is constructed based on sampling data. All current resolution elements are automatically associated with different classes in accordance with the common set of reference spectra and information measures of variability of these spectra. Parameters characterising the status of the objects of the "plant" class are restored for each resolution element of such objects.

EFFECT: high accuracy.

2 cl, 3 dwg