Radiotekhnika
Publishing house Radiotekhnika

"Publishing house Radiotekhnika":
scientific and technical literature.
Books and journals of publishing houses: IPRZHR, RS-PRESS, SCIENCE-PRESS


Тел.: +7 (495) 625-9241

 

Artificial neural network method operative mapping using unmanned aerial vehicles

Keywords:

M.V. Akinin – Ph. D. (Eng.), Junior Research Scientist, RSREU (Ryazan). E-mail: akinin.m.v@gmail.com N.V. Akinina – Post-graduate Student, RSREU (Ryazan). E-mail: natalya.akinina@gmail.com M.B. Nikiforov – Ph. D. (Eng.), Associate Professor, RSREU (Ryazan). E-mail: nikiforov.m.b@evm.rsreu.ru A.V. Sokolova – Undergraduate, RSREU (Ryazan). E-mail: alexandra.sokolova00@mail.ru A.I. Taganov – Dr. Sc. (Eng.), Professor, Head of Department, RSREU (Ryazan). E-mail: alxtag@yandex.ru


Currently, for many different socio-economic, environmental, and other problems applied vector maps with the information of the height of artificial and natural objects. In this article the method of surgical precision detailed mapping based on the use of neural networks and neural structures of various types, as well as on the use of unmanned aerial vehicles (UAVs) to carry out aerial photography in the visi-ble spectral range. In the first step of the mapping process of data is received from the unmanned aircraft. At the second stage processes the frames of a video sequence obtained using UAVs should be allocated for objects with multidimensional neural Kohonen’s maps. The next step is to find correspondences between the images allocated to the frame, and objects present on the digital terrain map. In the fourth stage of the images that have not been delivered to one correspondence objects on the map, required classification using tree classifier based on the use of support vector machines. As part of the development of software and hardware stand there have been several sessions of flight over the Nutwood lake (Ryazan region, Ryazan), during which were withdrawn following typical scene: a flat surface; almost flat surface with irregular small height dif-ference (mounds, ditches, pits); almost flat surface with grass growing on it; an aintenance shaft; mound (sharp slope); the water ob-ject; power lines. The results of this session flights were used for experimental studies operative mapping method. According to the results of experimental studies developed a method of mapping showed high accuracy (average - offset up to 2 meters from the actual position of objects, data processing time - no more than 1.35 seconds per frame).
References:

 

  1. Widyawardana Adiprawita, Adang Suwandi Ahmad, Jaka Semibiring Unmanned Aerial Vehicle Instrumentation for Rapid Aerial Photo System // International Conference on Intelligent Unmanned Systems. Bali, Indonesia. 2007. P. 111−118.
  2. Reid A., Ramos F., Sukkarieh S. Multi-Class Classification of Vegetation in Natural Environments Using an Unmanned Aerial System // International Conference on Robotics and Automation. China. 2011. P. 2953−2959.
  3. Haykin S. Neural networks and learning machines. Prentice Hall. 2009. 906 p.
  4. Sokolova A.V., Akinin M.V., Nikiforov M.B. Algoritm antialiasinga tekstur virtualnojj modeli mestnosti, osnovannyjj na metodakh debljuringa izobrazhenijj // Mezinarodni vedecka a prakticka konference «World & Science». CHekhija, Brno. 2014. P. 232−237.
  5. Fan R.-E., Chen P.-H., Lin C.-J. Working Set Selection Using Second Order Information for Training Support Vector Machines // Journal of Machine Learning Research (6). 2005. P. 1889−1918.
  6. Kohonen T. Self-organizing formation of topologically correct feature maps // Biol. Cyb. 1982. P. 59−69.
  7. Akinin M., Nikiforov M., Taganov A. Automated Combining Remote Sensing Data and Digital Maps, Based on Non-linear Cellular Neural Network and Boltzmann’s Restricted Stochastic Machine // 3rd Mediterranean Conference on Embedded Computing (MECO-2014). Montenegro, Budva. 2014. P. 144−147.
  8. Haralick M.R., Shanmugam K., Dinstein I. Textural features for image classification // IEEE Transactions on systems, man and cybernetics. 1973. V. SMC 3. № 6. P. 610−621.
  9. Akinina N.V., Akinin M.V., Nikiforov M.B. Primenenie nejjrosetevykh metodov analiza glavnykh komponent pri reshenii zadach obrabotki dannykh distancionnogo zondirovanija Zemli // Mezinarodni vedecka a prakticka konference «World & Science». CHekhija, Brno. 2014. P. 177−183.
  10. Akinin M.V., Konkin Y.V., Nikiforov M.B. Using Kitano\'s grammar encoding for finding optimal multilayer artificial neural network without feedback in image processing problems // Science and Education: Materials of the III International research and practice conference (V. I). Germany, Munich: Vela Verlag Waldkraiburg. 2013. P. 82−88.

 

© Издательство «РАДИОТЕХНИКА», 2004-2017            Тел.: (495) 625-9241                   Designed by [SWAP]Studio