Abstract

With the rapid development of remote sensing data acquisition and storage methods, today, a big data need to be processed and analyzed. These data are so large that the capabilities of the experts are no longer sufficient, leading to the continual increase in demand for automated data analysis methods. Application of artificial neural networks is one of the solutions to solve the above problem. The paper proposes a detailed algorithm for training multi-layered neural networks and remote sensing image classification application. Experimental results and evaluation of accuracy show that this method is quite feasible and simple to semi-automate the classification of objects represented on remote sensing images, for the establishment and updating of data national geospatial data.