Information technology for remotely sensed data processing

摘要

Proposed information technology for remotely sensed data processing is based on anomaly detection methods and algorithms, which are invariant to signal and background scale and structure, is advancement of existing approaches. The information technology combines the following conceptions and approaches: satellite images fractal self-similarity; multiresolution analysis of functional spaces on the base of wavelet transform (WT); using recursive fractal scanning for reduction of multidimensional spaces into one-dimensional segment; using nonparametric statistics for signal detection in noisy background. Proposed technology allows to: decrease the threshold of hidden signal detection by several times; detect the signals with unknown form; solve the problems connected to incomplete knowledge of signal, nonstationary, non-markovian and singular noises; improve the performance of subpixel signal detection and estimation in case of limited spatial and temporal resolutions of sensors. The investigation of Hausdorf dimension and index of Hoerst statistic is performed. Solution of the following problems is planned as extension of this research: conjugation between WT and recursive scannings for obtaining the powerful signal analysis mathematics, development of S/N ratio increase methods on the base of statistical image segmentation using nonparametric statistics; improvement of accuracy of subpixel signal estimation on the base of using the fractal properties of images.