The neuronet technology for aerospace monitoring data interpretation

For solving many practically important problems of ecology and landscape studying the information obtained by remote sensing (RS) methods plays an increasing role. Nowadays the development of high-automated methods and means of processing and interpretation of RS data is an extremely urgent problem. In the situations of training information lack and considerable uncertainty the most efficient approach for solving problems of RS data interpretation is application of neuronet algorithms of recognition of objects on images without training. The authors propose a neuronet technology for interpretation of aerospace monitoring data with the use of Kohonen's algorithm, based on a concept of dynamic kernels. The description of the proposed neuronet technology is given, particularities of its implementation are considered, and first results of application of the technology for solving problems of forest type mapping and assessing the pollution of reservoirs in the Tomsk region are shown.