V. A. Gorokhovatsky, Y. P. Putyatin, V. S. Stolyarov


Сontext. Increasing of productivity and extension of the functionality of modern computer vision systems require more effective methods for visual information processing. Main goals of structural recognition are related with the improvement of information classification technology in the space of features in a form of image key point descriptors, as well as the necessity of recognition performance estimation for application datasets. Particular attention is related to the investigation of data structure for the set of descriptors that directly affects the functioning of the recognition system.

Objective. Investigation of cluster representation for the set of structural features of application dataset was performed as well as the evaluation of cluster model performance in methods of visual objects structural recognition to provide compact representation of data was proposed.

Method. Methods of recognition based on transformation of structural features space by clustering and usage of cluster dataset image features were proposed. First method uses the integral representation of etalon images descriptions, the second one is based on the value of statistical distribution vector in matrix space cluster model during building the association between structural element and class. Result of research is creation of recognition methods and data models during construction of relevance vectors and features of classes in the transformed feature space.

Results. Using cluster transformation of the space of structural features allows to reduce the amount of computational costs, and improves recognition performance preserving desired efficiency in a hundred of times. Comparison between SURF and ORB methods for the formation of structural features was performed, processing time by ORB has appeared to be 60 times less. On the other hand, the set of SURF descriptors closely reflects the shape of visual objects. Modeling and experimental investigations of proposed recognition method for application dataset was performed. Effectiveness of the method in terms of efficiency was confirmed, comparative estimations of recognition quality depending on the level of additive noise for the analyzed treatment options were obtained.

Conclusions. Paper proposed the systematization and obtaining of perspective properties of recognition systems in the space of structural features of images. Classification methods based on cluster descriptions provide a sufficient level of image discrimination and high noise immunity. Scientific novelty of the research consists of synthesis of a method of structural image recognition based on the use of cluster processing and the construction of classification decisions in space of etalon cluster. Conversion to the vector-cluster presentation allows to significantly increase the speed of recognition by processing simplification. Practical value of paper is the obtaining of application program models for the modifications of structural image recognition method with the confirmation of the effectiveness and noise immunity of the proposed approach in a specific image dataset.


Computer vision; structural image recognition; set of structural features; SURF descriptors; ORB descriptors; relevance of descriptions; vector of characteristics of classes; noise immunity; recognition performance.


Gorohovatskij V. A. Strukturnyj analiz i intellektual’naja obrabotka dannyh v komp’juternom zrenii. H. Kompanija SMIT, 2014, 316 p.

Gorohovatskij V. A., Dunaevskaja M. D., Strunenko V. A. Izuchenie svojstv metodov klasterizacii primenitel’no k mnozhestvam harakternyh priznakov izobrazhenij, Sistemi obrobki іnformacії, 2016,Vip. 5 (142), pp. 124–127.

Bay H., Tuytelaars T., Van L. Gool Surf: Speeded up robust features, Computer Vision : Ninth European Conference on Computer Vision, Graz, 7–13 May, 2006: proceedings. Berlin, Springer, 2006, pp. 404–417.

Gorokhovatsky V. A. Efficient Estimation of Visual Object Relevance during Recognition through their Vector Descriptions, Telecommunications and Radio Engineering, 2016, Vol. 75, No. 14, pp. 1271–1283.

Rublee E., Rabaud V., Konolige K., Bradski G. ORB: an efficient alternative to SIFT or SURF, IEEE International Conference on Computer Vision (ICCV), November 06G.13, 2011, pp. 2564–2571. Rezhim dostupa

Gorohovatskij V. A., Gorohovatskij A. V., Berestovskij A. E. Strukturnoe raspoznavanie izobrazhenij s primeneniem modelej intellektual’noj obrabotki i samoorganizacii priznakov, Radio Electronics, Computer Science, Control, 2016, No. 3 (38), pp. 39–46.

Karami E., Prasad S., Shehata M. Image Matching Using SIFT, SURF, BRIEF and ORB: Performance Comparison for Distorted Images. Rezhim dostupa

Szeliski R. Computer Vision: Algorithms and Applications. London, Springer, 2010, 979 p.

Duda R. O., Hart P. E., Stork D. G. Pattern classification, 2ed. Wiley, 2000, 738 p.

Manning C. D., Raghavan P., Schutze H. Introduction to Information Retrieval, Cambridge, University Press, 2008, 528p.

Shapiro L. and Stockman G. Computer vision. Prentice Hall, 2001, 625 p.

GOST Style Citations

1. Гороховатский В. А. Структурный анализ и интеллектуальная обработка данных в компьютерном зрении / В. А. Гороховатский. – Х. : Компания СМИТ, 2014. – 316 с.

2. Гороховатский В. А. Изучение свойств методов кластеризации применительно к множествам характерных признаков изображений / В. А. Гороховатский, М. Д. Дунаевская, В. А. Струненко // Системи обробки інформації. – 2016. – Вип. 5 (142). – С. 124–127.

3. Bay H. Surf: Speeded up robust features / H. Bay, T. Tuytelaars, L.Van Gool // Computer Vision: Ninth European Conference on Computer Vision, Graz, 7–13 May, 2006: proceedings. – Berlin: Springer, 2006. – P. 404–417.

4. Gorokhovatsky V. A. Efficient Estimation of Visual Object Relevance during Recognition through their Vector Descriptions / V. A. Gorokhovatsky // Telecommunications and Radio Engineering. – 2016. – Vol. 75, No. 14. – P. 1271–1283.

5. Rublee E. ORB: an efficient alternative to SIFT or SURF / [E. Rublee, V. Rabaud, K. Konolige, G. Bradski] // IEEE International Conference on Computer Vision (ICCV), November 06–13, 2011, P. 2564–2571. Режим доступа

6. Гороховатский В. А. Структурное распознавание изображений с применением моделей интеллектуальной обработки и самоорганизации признаков / В. А. Гороховатский, А. В. Гороховатский, А. Е. Берестовский // Радиоэлектроника, информатика, управление. – 2016. – №3 (38). – C. 39–46.

7. Karami E. Image Matching Using SIFT, SURF, BRIEF and ORB: Performance Comparison for Distorted Images / E. Karami, S. Prasad, M. Shehata. – Режим доступа

8. Szeliski R. Computer Vision: Algorithms and Applications / R. Szeliski. – London : Springer, 2010. – 979 p.

9. Duda R.O. Pattern classification / R. O. Duda, P. E. Hart, D. G. Stork. – 2ed., Wiley, 2000. – 738 p.

10.Manning C. D. Introduction to Information Retrieval / C. D. Manning, P. Raghavan, H. Schutze. – Cambridge, University Press, 2008. – 528 p.

11. Shapiro L. Computer vision / L. Shapiro and G. Stockman. – Prentice Hall, 2001. – 625 p.


Copyright (c) 2017 V. A. Gorokhovatsky, Y. P. Putyatin, V. S. Stolyarov

Creative Commons License
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

Address of the journal editorial office:
Editorial office of the journal «Radio Electronics, Computer Science, Control»,
Zaporizhzhya National Technical University, 
Zhukovskiy street, 64, Zaporizhzhya, 69063, Ukraine. 
Telephone: +38-061-769-82-96 – the Editing and Publishing Department.

The reference to the journal is obligatory in the cases of complete or partial use of its materials.