Stankevich S. Method for Hyperspectral Imagery Informativity Evaluation in Earth Remote Sensing

Thesis for the degree of Doctor of Science in Engineering, specialty 05.07.12 - Remote Aerospace Research.- Scientific Centre for Aerospace Research of the Earth IGS NAS of Ukraine.- Kiev, 2008. The scientific problem of insufficient completeness and reliability of hyperspectral aerospace imagery informativity evaluation in Earth remote sensing tasks is solved. The new statistical spatial-frequency measure for hyperspectral imagery informativity is justified. These measure's components are the Kullback-Leibler information divergence, the equivalent spatial resolution and the equivalent "signal-to-noise" ratio. The multidimensional informational model and the statistical spatial-frequency one for hyperspectral imagery are developed. The new method for informativity evaluation and the optimization improved technique with maximum informativity criterion are developed too on this basis. For the first time the concept of the equivalent spatial resolution which describes the spatial-frequency properties of hyperspectral imagery using equivalence of object detection probabilistic performance is introduced. The new effective algorithms family for hyperspectral aerospace digital imagery statistical spatial-frequency processing is developed. The guidelines on implementation of the developed methods and algorithms in remote sensing practice are defined. The general methodology for hyperspectral aerospace imagery statistical spatial-frequency processing is formulated. This one determines principles and approaches to develop the new same methods using the statistical spatial-frequency paradigm. This research results provides the validity and efficiency increasing of the natural resourcing, scientific and special important thematic tasks in Earth remote sensing by objective informativity evaluation and optimization of hyperspectral aerospace imagery and also by essential reduction of computational requirements.