An Unsupervised Machine Learning Approach in Remote Sensing Data

The analysis, image recognition, the classification in remote sensing data always up to date is for the insiders a very useful system for the management, planning, protection of our territory. Technological evolution the succession of applications and tools in GIS environment, have refined more and more techniques allowing us to computerize much of the territory bringing an enrichment of geographical data distributed more or less updated, remote sensing, digital orthoimage help us in the areas where the local computerization is not always updated. In this paper I want to demonstrate that an approach based on machine learning applied to territorial analysis using dynamic matrices, provides for the definition of numerical indices assigned to each recognized pattern. The numerical indices will be the elements of recognition where the territorial computerization is insufficient. The data from remote sensing or digital orthoimage are variable matrices where the numerical indices have to express their potential. The goal is to educate the recognition in matrix form to obtain a dynamic topology where the computerization has not been updated.