Algorithm for Object Image Automatic Tracking

Development of reliable and precise object image selection algorithms, which are invariant to object image texture, background and noise parameters, and moreover provide real-time image processing, is a relevant problem. Random Ferns method seems suitable for developing algorithms with similar qualities. In this paper, we propose an object image processing algorithm, which provides object image selection and determination of the position data and overall dimensions for the selected object image during the object tracking. The proposed algorithm is based on Random Ferns method, and requires initial training. We carried out numerical experiments to evaluate the efficiency of such algorithm for several types of objects, whose images, with complicated image texture, were exposed on the spatially-irregular background. Moreover, we blurred images with white Gaussian noise in order to vary the signal-to-noise ratio. Findings of the research show that the object image selection algorithm, based on Random Ferns method, in case if the signal-to-noise ratio is greater than 10, provides reliable object image selection, i. e. correct selection probability is near to 100 %. The algorithm provides position data and overall dimensions measurement error less than three pixels. The algorithm provides high performance. The average computation time didn't exceed 5 ms even for selecting image of the object with complicated texture, which is exposed on spatially-irregular background. Consequently, it is suitable for real-time working computer vision systems

References

[1] Alpatov B.A., Babayan P.V., Balashov O.E., Stepashkin A.I. Metody avtomaticheskogo obnaruzheniya i soprovozhdeniya ob"ektov. Obrabotka izobrazheniy i upravlenie [Automatic methods of objects acquisition and tracking. Image processing and control]. Moscow, Radiotekhnika Publ., 2008. 176 p.

[2] Alpatov B.A., Babayan P.V. Imaging processing and analysis methods in airborne objects acquisition and tracking systems. Tsifrovaya obrabotka signalov [Digital Signal Processing], 2006, no. 2, pp. 45−51 (in Russ.).

[3] Bakut P.A., ed. Using image segmentation methods in autonomous systems of moving object acquisition, recognition and tracking system. Zarubezhnaya radioelektronika, 1987, no. 10, pp. 3−93 (in Russ.).

[4] Alpatov B.A., Blokhin A.N., Murav'yev V.S. Image processing algorithm for aerial objects automatic tracking system. Tsifrovaya obrabotka signalov [Digital Signal Processing], 2010, no. 4, pp. 12−17 (in Russ.).

[5] Murav'yev V.S., Murav'yev S.I. An adaptive algorithm of aerial object extraction and detection on images. Informatsionno-upravlyayushchie sistemy [Information and Control Systems], 2011, no. 5, pp. 8−14 (in Russ.).

[6] Kalal Z., Mikolajczyk K., Matas J. Tracking-learning-detection. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2012, vol. 34, no. 7, pp. 1409−1422. DOI: 10.1109/TPAMI.2011.239

[7] [1] Byshevskaya-Konopko L.O., Vorob'yev I.L., Izyneev A.A., Sadovskiy P.I. Heat dissipation accompanying lamp pumping in an erbium active element. Journal of Optical Technology, 2013, vol. 80, no. 10, pp. 594–600. DOI: 10.1364/JOT.80.000594 Available at: http://itas2015.iitp.ru/pdf/1570177051.pdf (accessed: 07.12.2015).

[8] Boykov V.A., Kolyuchkin V.Ya. Algorithm for object image selection on a randomized spatially-irregular background. Informatsionno-izmeritelnye i upravlyayushchie sistemy, 2016, no. 3, pp. 29−36 (in Russ.).

[9] Ozuysal M., Calonder M., Lepetit V., Fua P. Fast keypoint recognition using random ferns. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2010, vol. 32, no. 3, pp. 448−461. DOI: 10.1109/TPAMI.2009.23

[10] Calonder M., Lepetit V., Fua P. BRIEF: Binary robust independent elementary features. Computer Vision — ECCV’10, 2010, pp. 778−792.