• 1990 (Vol.4)
  • 1989 (Vol.3)
  • 1988 (Vol.2)
  • 1987 (Vol.1)

Volume 30 #3

Contents

  1. The structural-functional organization of mammalian auditory cortex as the basis of acoustic information cortical processing. The structural organization
  2. Allocation of the characteristic features of in low-frequency envelope of the tone by the neurons in the midbrain auditory center of the frog
  3. Discrimination of sound signals with rippled spectra in the noises of different spectral compositions
  4. Temporary threshold shifts in naïve and experienced belugas
  5. The role of simpatic-epinephrine system in the changes of electrical activity of the brain hemisphere cortex and hypothalamus under vibration exposure
  6. Cyclic polypeptide PP-14 modulates the voltage sensitivity of slow sodium channels
  7. Training optimal Viola–Jones detectors using greedy algorithms for selecting control parameters with intermediate validation on each level
  8. Abnormal band in the circular dichroism spectrum of DNA cholesteric liquid-crystalline dispersions – an analytical criteriom for detection of coloured biologically active compounds

Training optimal Viola–Jones detectors using greedy algorithms for selecting control parameters with intermediate validation on each level

© 2016 I. V. Polyakov, E. G. Kuznetsova, S. A. Usilin, D. P. Nikolaev

Institute for Information Transmission Problems RAS 127051 Moscow, Bolshoi Karetny lane, 19
Federal Research Center “Computer Science and Control” of RAS 119333 Moscow, Vavilova str., 44/2
Moscow Institute of Physics and Technology (state university) 141700 Moscow Region, Dolgoprudny, Institutsky lane, 9

Received 18 Apr 2016

Nowadays object detection using video stream problem is considered solved. One of the most popular approaches to cope with this problem is Viola-Jones cascade learning. However our detection problem is appear to be highly complicated in practice due to variation of object form and lightness conditions in scene. In that case there are two ways to improve detection quality: trying to find another method or learning method improvement. In this paper we consider Viola–Jones learning parameter selection to optimize resulting detector in “Selectivity – Specificity – Productivity” space. We investigate greedy algorithm for selection learning parameters on each level. Also we state that it is possible to significantly improve cascade quality using intermediate level-wise checking under validation set. We provide paper by experimental results.

Key words: Viola–Jones, machine learning, object detection, control parameters optimization, recognition, uncontrolled conditions, video stream recognition

Cite: Polyakov I. V., Kuznetsova E. G., Usilin S. A., Nikolaev D. P. Postroenie optimalnykh kaskadov violy–dzhonsa pri pomoshchi “zhadnykh” algoritmov perebora upravlyayushchikh parametrov s promezhutochnym kontrolem po validatsionnoi vyborke [Training optimal viola–jones detectors using greedy algorithms for selecting control parameters with intermediate validation on each level]. Sensornye sistemy [Sensory systems]. 2016. V. 30(3). P. 241-248 (in Russian).

References:

  • Zherdev D.A., Kazanskiy N.L., Fursov V.A. Object recognition in radar images using conjugation indices and support subspaces // Comp. opt. 2015. V. 39. No 2. p. 255–264 [in Russian].
  • Minkina A.G., Grigoriev A.S., Usilin S.A., Polevoi D.V., Nikolaev D.P. Generalization of the Viola–Jones method as a decision tree of strong classifiers for real-time object recognition in video stream // Inform. Techn. Sys. 2014. P. 158–163 [in Russian].
  • Usilin S.A., Nikolaev D.P., Sholomov D.L. Arlazarov V.V. Guilloche Elements Recognition Applied to Passport Page Processing // The Proceedings of Institute for Systems Analysis of Russian Academy of Sciences. Processing of information and graphic resources. 2013. V. 63. P. 106–110 [in Russian].
  • Usilin S.A., Nikolaev D.P., Postnikov V.V. Localization, orientation and documents identification with a fixed image geometry // The Proceedings of Institute for Systems Analysis of RAS. Proc. inform. graphic resources. 2010. P. 248–261 [in Russian].
  • Chochia P., Milukova O. Comparison of two-dimensional variations in the context of the digital image complexity Assessment // J. Communicat. Technol. Electronics. 2015. V. 60. No 12. P. 1432–1440.
  • Chiu C.C., Ku M. Y., Wang C.Y. Traffic surveillance system for Vision-Based vehicle recognition and tracking // J. Inform. Sc. Engin. 2010. V. 26. P. 611–629.
  • Freund Y., Schapire R. A decision-theoretic generalization of on-line learning and an application to boosting // J. Comp. System Sciences. 1997. V. 55. P. 119–139.
  • Karpenko S., Konovalenko I., Miller A., Miller B., Nikolaev D. Visual navigation of the UAVs on the basis of 3D natural landmarks // Proc. SPIE. 8th Internat. Conf. Machine Vision. 2015. V. 9875. P. 1–10.
  • Konovalenko I., Miller A., Miller B., Nikolaev D. UAV navigation on the basis of the feature points detection on underlying surface // Proc. 29th Europ. Conf. Modell. Simulat. 2015. P. 499–505.
  • Kuznetsova E., Shvets E., Nikolaev D. Viola–Jones based hybrid framework for real-time object detection in multispectral images // Proc. SPIE. Eighth Internat. Conf. Machine Vision. 2015. V. 9875. P. 1–6.
  • Minkina A., Nikolaev D., Usilin S., KozyrevV.
  • Generalization of the Viola–Jones method as a decision tree of strong classifiers for real-time object recognition in video stream // Proc. SPIE. 7th Internat. Conf. Machine Vision. 2015. V. 9445. P. 1–5.
  • Viola P., Jones M.J. Robust real-time face detection // Intern. J. Comp. Vision. 2004. V. 57. P. 137–154.
  • Yakimov P.Yu. Preprocessing digital Images for quickly and reliably detecting road signs // Pattern Recogn. Image Analysis. 2015. V. 25(4). P. 729–732.