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

 

Efficiency of stochastic gradient object identification procedure for binary images

Keywords:

R.G. Magdeev – post-graduate student of Ulyanovsk State Technical University. E-mail: radiktkd2@yandex.ru
A.G. Tashlinskii – Dr. Sci. (Eng.), Professor, head of Radio engineering department, Ulyanovsk State Technical University. E-mail: tag@ulstu.ru


A comparative analysis of correlation-extreme algorithm (CEA), contour analysis method (CAM) and stochastic gradient (SGI) object identification methods with respect to computational cost and incorrect identification probability is performed. It is shown that CAM has the lowest cost with small images, with big images – SGI. CEA has quadratic dependence of the cost on the image size both in spatial and frequency domains and it is 100 times higher. The computational cost of CAM, SGI and CEA is slightly dependent on the object size in frequency domain, but in spatial domain for CEA it has a quadratic form. SGI requires the lowest computational cost, CEA – the highest. Due to the largest sample size CEA in spatial domain has the highest noise stability. CEA has some incorrect identifications due to step values used for identification parameters. SGI has also shown good noise stability. But the probability of the correct identification for this method is dependent on misalignment between tested and standard objects. The probability of incorrect identification for CAM is several times higher is case of noise due to errors in contour detection. Parametric optimization possibilities of the methods are considered. The values to optimize for CEA are steps of object location parameters determining the speed and accuracy. For CAM the length of the contour-vector is an important parameter. The number of different parameters’ initial approximations of the reference standard of the same type depending on the method workspace determines the computational cost of SGI. It is also shown that the use of inter-frame correlation coefficient as the objective function provides bigger parameters’ domain in comparison with the mean square difference.
References:

  1. Vvedenie v konturny'j analiz / pod red. Ja.A. Furmana. M.: Fizmatlit. 2003. 592 s.
  2. Gonsales R., Vuds R. Cifrovaya obrabotka izobrazhenij. M.: Texnosfera. 2005. 1072 s.
  3. Pre'tt U. Cifrovaya obrabotka izobrazhenij: Per. s angl. M.: Mir. 1982. 480 s.
  4. Gruzman I.S., Kirichuk V.S., Kosy'x V.P., Peretyagin G.I., Spektor A.A. Cifrovaya obrabotka izobrazhenij v informaczionny'x sistemax: Uchebnik NGTU: uchebnoe posobie. Novosibirsk: Izd-vo NGTU. 2002. 352 c.
  5. Cy'pkin Ja.Z. Informaczionnaya teoriya identifikaczii. M.: Nauka. Fizmatlit. 1995. 336 s.
  6. Tashlinskij A.G. Psevdogradientnoe oczenivanie prostranstvenny'x deformaczij posledovatel'nosti izobrazhenij // Naukoemkie texnologii. 2002. T. 3. S. 32-43.
  7. Canny J. A computational approach to edge detection. Pattern Analysis and Machine Intelligence // IEEE Transactions on, PAMI-8(6). 679–698. Nov. 1986.
  8. Duda R., Xart P. Raspoznavanie obrazov i analiz sczen. M.: Mir. 1976. 511 s.
  9. Tashlinskii A.G. The specifics of pseudogradient estimation of geometric deformations in image sequences // Pattern Recognition and Image Analysis. 2008. V. 18. № 4. P. 700-705.
  10. Tashlinskij A.G. Psevdogradientnoe oczenivanie prostranstvenny'x deformaczij posledovatel'nosti izobrazhenij // Naukoemkie texnologii. 2002. T. 3. S. 32-43.
  11. Tashlinskii A.G., Smirnov P.V., Zhukov S.S. Analysis of Methods of Estimating Objective Function Gradient during Recurrent Measurements of Image Parameters // Pattern Recognition and Image Analysis. 2012. V. 22. № 2. P. 393-399.
  12. Kruchinin A.Ju. E'ffektivnost' upravleniya proczessom raspoznavaniya obrazov v real'nom vremeni // Trudy' IX mezhdunarodnoj nauchno-texnicheskoj konferenczii «Novy'e informaczionny'e texnologii i sistemy'». Penza: Izd-vo PGU. 2010. Ch. 1. S. 150-155.
  13. Sebryakov G.G., Soshnikov V.N., Kikin I.S., Ishutin A.A. Optimizacziya parametrov razbieniya analiziruemogo fragmenta izobrazheniya sczeny' po kriteriyam kachestva i vy'chislitel'noj e'ffektivnosti raspoznavaniya nablyudaemy'x ob''ektov // Tezisy' dokladov nauchno-texnicheskoj konferenczii «Texnicheskoe zrenie v sistemax upravleniya – 2014». M.: Institut kosmicheskix issledovanij RAN. S. 149-151.
  14. Tashlinskii A.G., Dikarina G.V., Minkina G.L., Repin A.N. Pseudogradient optimization in the estimation of geometric interframe image deformations // Pattern Recognition and Image Analysis. 2008. V. 18. № 4. P. 706-711.
  15. Tashlinskii A.G., Muratkhanov D.S. Structural optimization of pseudogradient algorithms for measuring interframe image deformations // Pattern Recognition and Image Analysis. 2003. V. 13. № 1. P. 177-178.

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