Ways to improve objects recognition and classification in satellite images

Abstract

The article analyzes modern technologies and methods to improve the efficiency of object recognition in satellite images, in particular the method of multispectral satellite high-resolution scans and their interpretation in a geographical information system (GIS).Using multispectral images can improve the efficiency of objects recognition and classification. However, at a sufficiently high spectral resolution there is a problem related to the necessity of characteristics (spectral signatures) processing in high-dimensional spaces. The solution to this problem lies in the fact that first it is reasonable to reduce the space dimension and to perform recognition (classification) in the new space. Increasing of the separation ability resolves two interrelated objectives: improving of the visual quality and images reconstruction. Solution of the first problem is the method of fragmentation and zoning images. The solution of the second one is the deconvolution method.The combination of area images processing and their reconstruction allow approaching solution of fire prediction problem and selection of distinguishing methods.