Sovremennye problemy distantsionnogo zondirovaniya Zemli iz kosmosa, 2009. Issue.6. Vol.1. P. 196-205
Natural and anthropogenic objects pattern recognition and their
condition assessment using multispectral and hyperspectral airspace
remote sensing data
В.В. Козодеров1, В.В. Борзяк1, Е.Д. Дмитриев2, В.Д. Егоров2
1 Московский государственный университет им. М.В.Ломоносова
2 Институт вычислительной математики РАН
Examples are considered of the previously elaborated algorithm and software tools adaptation
concerning multispectral space imagery processing to develop applications on the soil-vegetation cover
objects pattern recognition using the hyperspectral instrument in-flight testing results. New opportunities
are open up for the thin structure analysis of the obtained spectra with spectral resolution within a
nanometer unit to solve problems of the remotely sensed nano-diagnostics of the observed objects. The
pattern recognition computational procedures are based on finding a Euclidean distance minimum in a
pattern space for the relevant objects relative to some specific objects to be selected during georeferencing
of particular pixel on the hyperspectral images to the synchronized data of the area air photo
survey. Comparison is given of the gained results with data of ground forest taxation observations for the
selected area. A conclusion is made about an urgent necessity to collect the accurate data about the forest
density parameters within separate plots to account for reliability of the obtained results.
Keywords: natural and anthropogenic objects, multispectral and hyperspectral imagery processing, pattern recognition, computational mathematics techniques, ground forest taxation observations
Full text