ALGORITM SCENE RECOGNITION FOR AUGMENTED REALITY SYSTEMS 

Finogeev Aleksey Germanovich, Doctor of engineering sciences, professor, sub-department of CAD, Penza State University (Penza, 40 Krasnaya str.), finogeev@sura.ru
Chetvergova Mariya Vladimirovna, Postgraduate student, Penza State University (Penza, 40 Krasnaya str.), marysha85@inbox.ru 

004.932:159.9 

The article considers the technology of critical point recognition in object images taken from a video portion for a mobile system of augmented reality. The authors suggest of determining and matching the descriptors of critical points on the object images with further possibility of recognition system learning on the basis of using a forest of random trees. The study compares the existing methods point feature recognition, such as SIFT, SURF and RIFF, and describes their drawbacks. In order to increase the quality of the method of recognizing the objects with isolated point features on mobile devices the researchers suggest a method on the basis of a forest of random trees. The general idea of the method is recognition of objects on the basis of Bayes classifier distribution statistics on possible descriptors matching. There are two structures of random tree forests suggested: basic and augmented. The authors conducted comparative analytical and experimental research of the existing and the developed methods of object recognition with detected point features on the basis of a forest of random trees. It is shown that the application of the developed method on the basis of random trees lead to a better recognition quality value in conditions of higher requirements to random-access memory and minimal increase of sytem work time. 

augmented reality, scene recognition, point feature, special point, detecting, descriptor, random tree. 

1. Vasil'ev V. N., Gurov I. P., Potapov A. S. Vserossiyskiy konkurs obzorno-analiticheskikh statey po prioritetnomu napravleniyu «Informatsionno-telekommunikatsionnye sistemy» [All-Russia competition of reviewing and analytical articles on the top-priority research field “Information-telecommunication systems”]. available at: http://www. itc.edu.ru/itkonkurs2008/
2. Shi J., Kanade T. IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR’94), IEEE Computer Society, Sietl. 1994.
3. Harris C., Stephens M. In Fourth Alvey Vision Conference. Manchester, UK, pp. 147– 151.
4. Lowe D. G. Proceedings of the International Conference on Computer Vision. 1999, vol. 2, pp. 1150–1157.
5. Herbert B., Ess Andreas, Tuytelaars Tinne, Luc Van Gool. Computer Vision and Image Understanding (CVIU). 2008, vol. 110, no. 3, pp. 346–359.
6. Takacs G., Chandrasekhar V., Tsai S., Chen D., Grzeszczuk R., Girod B. Unified Real- Time Tracking and Recognition with Rotation-Invariant Fast Features. available at: CVPR2010_RIFF.pdf
7. Calonder M., Lepetit V., Fua P. BRIEF: Binary Robust Independent Elementary Features. available at: http://cvlab.epfl.ch/~lepetit/
8. Hartley R., Zisserman A. Multiple View Geometry In Computer Vision. 2nd edition. Cambridge: Cambridge University Press, 2003. p. 670.