RECOGNITION OF NATURAL AND ANTROPOGENIC NETS BASED ON THE GRAPH THEORY AND METHODS OF MACHINE LEARNING

http://doi.org/10.24057/2414-9179-2018-2-24-340-346

View or download the article (Rus)

About the Authors

Anastasia A. Shurygina

Lomonosov Moscow State University,
Leninskie gory, 1, 119991, Moscow, Russia,
E-mail: shur.a17@yandex.ru

Timofey E. Samsonov

Lomonosov Moscow State University,
Leninskie gory, 1, 119991, Moscow, Russia,
E-mail: tsamsonov@geogr.msu.ru

Abstract

Current study is a part of the research which is aimed at identifying approaches to the recognition of spatial data objects presented in the vector data models. In the previous stage the classification of spatial objects was based on the set of their morphometric features—shape characteristics. A conclusion was drawn that such attributes were insufficient for automated recognition and there was a need in additional study of the spatial relationships of objects. That means the transition from the object recognition level to the feature classes recognition level, from shape analysis to the exploraton of objects combinations—patterns. It is important to analyse spatial relationships between objects of the same classes as well as different ones. That study covers the problem of classifying various feature classes.

Application options of the study results are described in the paper on a par with the methodology of solving the task and materials involved in the study. Moreover, the results of objecs recognition based on their morphometric features and graph characteristics are compared.

The paper presents the ways of automated recognition of natural and antropogenic vector objects with the linear localization. The study applies Graph theory and Machine learning to classify them. The Python script for calculating graph parameters of linear objects was created. Furthermore, the model in Rapid Miner Studio application program was prepared for spatial objects’ nets recognition.

Keywords

graph theory, machine learning, pattern recognition.

References

  1. Lurie I.K., Kosikov A.G. Theory and practice of image processing. Remote sensing and geographic information systems / Edited by A.M. Berlant. М.: Nauchniy mir, 2003 (in Russian).
  2. Newman M.E.J. Mixing patterns in networks. Physical Review E. 2003. Т. 67, No 2.
  3. Reingold O. Undirected connectivity in log-space. Journal of the ACM (JACM). 2008. Т. 55, No 4. P. 17.
  4. Schank T., Wagner D. Approximating clustering-coefficient and transitivity. Universität Karlsruhe, Fakultät für Informatik, 2004.
  5. Shurygina A.A., Samsonov T.E. Analisys and systematization of morphometric characteristics of major classes of geographic maps’ objects. Scientific research of young scientists-cartographers performed under the guidance of the staff of the Department of Cartography and Geoinformatics of the Faculty of Geography of Lomonosov Moscow State University / Ed. by M.V. Zimin, P.E. Kargashin, M.V. Kusilman et al. KDU, Moscow, 2017. P. 110–121 (in Russian).

For citation: Shurygina A.A., Samsonov T.E. RECOGNITION OF NATURAL AND ANTROPOGENIC NETS BASED ON THE GRAPH THEORY AND METHODS OF MACHINE LEARNING Proceedings of the International conference “InterCarto. InterGIS”. 2018;24(2):340–346 http://doi.org/10.24057/2414-9179-2018-2-24-340-346