USE OF MACHINE LEARNING TECHNOLOGIES IN DECISION OF GEOINFORMATIONAL TASKS

http://doi.org/10.24057/2414-9179-2018-2-24-371-384

View or download the article (Rus)

About the Authors

Alexey A. Kolesnikov

Siberian State University of Geosystems and Technologies,
Plakhotnogo str., 10, 630108, Novosibirsk, Russia,
E-mail: alexeykw@mail.ru

Pavel M. Kikin

Siberian State University of Geosystems and Technologies,
Plakhotnogo str., 10, 630108, Novosibirsk, Russia,
E-mail: it-technologies@yandex.ru

Elena V. Komissarova

Siberian State University of Geosystems and Technologies,
Plakhotnogo str., 10, 630108, Novosibirsk, Russia,
E-mail: komissarova_e@mail.ru

Elena L. Kasyanova

Siberian State University of Geosystems and Technologies,
Plakhotnogo str., 10, 630108, Novosibirsk, Russia,
E-mail: helenkass@mail.ru

Abstract

Computer vision and data analysis are one of the most popular topics both among information technologies and most areas of scientific research. Geography, cartography and geoinformatics with their variety of types of source data, spatial analysis problems, visualization methods, modeling and forecasting methods cannot be better suited for using modern algorithms of machine learning. However, the use of these technologies rarely goes beyond the solution of private tasks of commercial campaigns and, often, is not widely publicized and any systematization or scientific description. In this respect, we decided to make a research of machine learning technologies in the context of using it while solving the most typical problems of geographical research. The classification of problems, algorithms and methods of computer vision from the point of view of geoinformation systems is given. Possible ways of solving some problems of classification and segmentation of raster images are described. The most popular of them are analyzed, including such as the use of convolutional and pre-conditioned neural networks for the recognition of objects on satellite images. The approbation took place within the competition in vectorization of hydrographic objects and the classification of objects in the open sea Statoil/C-CORE Iceberg Classifier Challenge. As initial data, we took marked satellite images of the water surface. The ways of spatial data analysis using the Moran index and calculating the Gini coefficient are considered. The methods of predicting the location of the coordinates of the house and work sudden user using the time series of ATM and cash register transactions at service points using regression algorithms were investigated. To conduct this study, the data set of the All-Russian competition in machine learning Raiffeisen Data Cup was used. We compared the results of usage of the machine learning algorithms and traditional methods of spatial analysis. Based on the results of the fulfilled investigations, we made the conclusions about the usability of the algorithms and technologies for specific geographic tasks, taking into account the dependence of the results from the types of using data used, resources requirements, accuracy, and universality.

Keywords

machine learning, cartography, segmentation, neural networks, regression, geoinformatics, aerial images.

References

  1. Badrinarayanan V., Kendall A., Cipolla R. Convolutional encoder-decoder architecture for image segmentation. IEEE Transactions on Pattern Analysis and Machine Intelligence, IEEE Transactions on Pattern Analysis and Machine Intelligence, 39 (12), No 7803544. 2017. P. 2481–2495.
  2. Benedetti R., Rossini P. On the use of NDVI profiles as a tool for agricultural statistics the case study of wheat yield estimate and forecast in Emilia Romanga. Remote Sensing of Environment. 1993. V. 45. P. 311–326.
  3. Bottou L. Large-scale machine learning with stochastic gradient descent. Proceedings of COMPSTAT’ 2010. Springer, 2010. P. 177–186.
  4. Breiman L. Random forests. Machine learning. 2001. Т. 45, No 1. P. 5–32.
  5. Brown F.J., Reed C.B., Hayes J.M. et al. A prototype drought monitoring system integrating climate and satellite data. Proceedings of the Pecora L5/land satellite information 1V/ISPRS commission I/FIEOS. 2002. Colarado, USA.
  6. Chen L.-C., Papandreou G., Kokkinos I. et al. Deeplab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected crfs. arXiv preprint arXiv:1606.00915, 2016.
  7. Dai J., He K., Sun J. Boxsup: Exploiting bounding boxes to supervise convolutional networks for semantic segmentation. Proceedings of the IEEE International Conference on Computer Vision. 2015. P. 1635–1643.
  8. Eigen D., Fergus R. Predicting depth, surface normals and semantic labels with a common multi-scale convolutional architecture. Proceedings of the IEEE International Conference on Computer Vision. 2015. P. 2650–2658; for Large-Scale Image Recognition, CoRR, vol. abs/1409.1556, 2014.
  9. Girshick R., Donahue J., Darrell T., Malik J. Rich feature hierarchies for accurate object detection and semantic segmentation. Proceedings of the IEEE conference on computer vision and pattern recognition. 2014. P. 580–587.
  10. Giusti J. Guzzi, Cires D.C., He F.-L. et al. A machine learning approach to visual perception of forest trails for mobile robots. IEEE Robotics and Automation Letters. 2016. V. 1, No 2. P. 661–667.
  11. Goodfellow I., Bengio Y., Courville A. Deep Learning. MIT Press. 2016. 800 p. ISBN: 9780262035613.
  12. Hariharan B., Arbelaez P., Girshick R., Malik J. Simultaneous detection and segmentation. European Conference on Computer Vision. Springer, 2014. P. 297–312.
  13. Haug S., Ostermann J. A Crop Weed Field Image Dataset for the Evaluation of Computer Vision Based Precision Agriculture Tasks. Computer Vision—ECCV 2014 Workshops. Zurich: Springer, 2014. P. 105–116.
  14. Hung C., Nieto J., Taylor Z. et al. Orchard fruit segmentation using multi-spectral feature learning. Intelligent Robots and Systems (IROS) IEEE/RSJ International Conference on. IEEE. 2013. P. 5314–5320.
  15. Jia Y., Shelhamer E., Donahue J. et al. Caffe: Convolutional architecture for fast feature embedding. arXiv preprint arXiv: 1408.5093, 2014.
  16. Kolesnikov A.A., Kikin P.M., Komissarova E.V. Using open semantic databases to obtain spatial information. Interekspo GEO-Sibir’-2017. XIII Mezhdunar. nauch. kongr., 17–21 aprelya 2017 g., Novosibirsk. Mezhdunar. nauch. konf. “Geodeziya, geoinformatika, kartografiya, marksheyderiya”: V. 2. Novosibirsk: SSUGT, 2017. P. 53–57 (in Russian).
  17. Kolesnikov A.A., Kikin P.M., Komissarova E.V. Creating virtual models of terrain and buildings. Interekspo GEO-Sibir’-2016. XII Mezhdunar. nauch. kongr., 18–22 aprelya 2016 g., Novosibirsk. Mezhdunar. nauch. Konf. “Raneye preduprezhdeniye i upravleniye v krizisnykh situatsiyakh v epokhu bol’shikh dannykh”: sb. materialov. V. 2. Novosibirsk: SSUGT, 2016. P. 37–40 (in Russian).
  18. Kolesnikov A.A., Kikin P.M., Komissarova E.V., Grishenko D.V. Using machine learning for mapping. Mezhdunarodnaya nauchno-prakticheskaya konferentsiya “Ot karty proshlogo—k karte budushchego”, 28–30 noyabrya 2017, g. Perm’—g. Kudymkar. P. 110–120 (in Russian).
  19. Kolesnikov A.A., Kikin P.M., Komissarova E.V. Programming for GIS in the framework of the preparation of students studying in the specialty “Information Systems and Technologies”. Aktual’nyye voprosy obrazovaniya. Vedushchaya rol’ univ-ta v tekhn. i kadrovoy modernizatsii rossiyskoy ekonomiki: sb. mater. Mezhdunar. nauchno-metodich. konf. Part 1. Novosibirsk: SSUGT, 2015. P. 87–90 (in Russian).
  20. Mitchell T. Machine learning. McGraw-Hill, 1997. 414 p. ISBN: 0070428077.
  21. Mortensen K., Dyrmann M., Karstoft H. et al. Semantic segmentation of mixed crops using deep convolutional neural network. International Conference on Agricultural Engineering, 2016.
  22. Peters J.A., Walter-Shea A.E., Ji L. et al. Drought monitoring with NDVI-based standardized vegetation index. Photogrammetric Engineering and Remote Sensing. 2002. V. 68: 7175.
  23. Rey S.J., Smith R.J. A spatial decomposition of the Gini coefficient. Letters in Spatial and Resource Sciences. 2013. V. 6. P. 55–70.
  24. Russakovsky O., Deng J., Su H. et al. ImageNet Large Scale Visual Recognition Challenge. International Journal of Computer Vision (IJCV). 2015. V. 115, No 3. P. 211–252.
  25. Simonyan K., Zisserman A. Very Deep Convolutional Networks for Large-Scale Image Recognition. arXiv technical report, eprint arXiv:1409.1556, 2014.

For citation: Kolesnikov A.A., Kikin P.M., Komissarova E.V., Kasyanova E.L. USE OF MACHINE LEARNING TECHNOLOGIES IN DECISION OF GEOINFORMATIONAL TASKS Proceedings of the International conference “InterCarto. InterGIS”. 2018;24(2):371–384 http://doi.org/10.24057/2414-9179-2018-2-24-371-384