Combined automated searching for coassociative relations as a preprocessing step in exploratory multispectral data analysis

In current research a combination of several techniques - group of cluster analysis approaches and machine learning methods have been used to investigate the satellite multispectral imagery of cropland. The primary work with data makes emphasis on clustering by a group of algorithms. The method proves to make the neural network detecting of pixels with similar signatures in much more accurate way and allows to interpret crop growth processes correctly. The technique appears useful in order to form a representative training set for powerful neural image classification model to provide the accuracy of revealing structural dependencies and carrying them over on new data. The theoretical part of research is given to plan further experimental research work.