Optimizing of the Number of Spectral Channels in Problems of Processing and Analysis of Hyperspectral Remote Sensing of the Ocean Data

Maltsev G. N., Kozinov I. A.

The number of modern channels of multispectral and hyperspectral modern optoelectronic systems for remote sensing is redundant for many tasks of the ocean monitoring and requires minimization. The aim of this optimization is a creation of a selection of images obtained from different most informative spectral channels. The selection in dimension is significantly lower than the number of channels of hyperspectral system and provides a solution to the problem of thematic processing. Mathematically, the problem of choosing the most informative spectral channels of hyperspectral survey for pixel of sensor is formulated as the problem of detecting changes in the properties of the registered coordinates or reference discrete spectral image described by a set of spectral components. Set of discrete spectral image components in the analysis is considered as a sequence of independent random Gaussian variables with variance and piecewise constant mean, which abruptly changes from one discrete location to another. The algorithm for solving this problem using methods of statistical estimation as a crucial statistic for detection and maximum likelihood estimation coordinates (spectral channel) changes in the properties of the analyzed process is shown. The consistent application of the algorithm to the sample values of the spectral components allows determination of numbers of the most informative spectral channels. The task of the configuration parameters of the synthesized algorithm to select the most informative spectral channels is considered. The basic parameters are the adjustable threshold value and the size of the sliding window. In accordance with the considered algorithms, implemented in MathLab software programming environment, an example of selecting the most informative spectral channels for spectral image obtained from hyperspectral image pixel of coastal area overgrown with algae is presented.

Download original text