STATISTICAL DISTRIBUTIONS AND CHAIN REPRESENTATION OF DATA WHEN DETERMINING THE RELEVANCE OF STRUCTURAL DESCRIPTIONS OF VISUAL OBJECTS
DOI:
https://doi.org/10.26906/SUNZ.2018.6.087Keywords:
structural image recognition methods, key point, BRISK detector, chain representation, fragment data distribution, general descriptor, descriptive relevance, voting, Manhattan metric, speed of relevancy estimationAbstract
The subjects of the paper are the models for estimation of the relevance between images in the space of key point descriptors when recognizing visual objects in computer vision systems. The goal is to create an image structural recognition method based on the implementation of chain data models using probability distributions of the sets of descriptors. The tasks include the development of mathematical and software models of efficient data analysis for determining the relevance of structural descriptions, investigation of the properties, application attributes, values of parameters of these models, evaluation of the effectiveness of the specific image processing. The methods are used: a BRISK detector for forming the key point descriptors, data mining, methods of bitwise processing and building bit-data distributions, a method of metric relevance estimation, software modeling. The following results were obtained. The transition from the sets of descriptors to probability distributions of fragments and the comparison of images in the space of distributions provide the necessary recognition performance. Data processing and analysis are performed hundreds of times faster than traditional vote counting. Processing and analysis of bit combinations forms significant properties for a set of description elements with keeping the data structure and their unification. With an increase of the number of bits in the distribution fragment, the distance between images increases and it contributes to an increase of their difference degree. The chain representation and the use of distributions create a new data space, which allows to improve the performance of image recognition systems significantly. Conclusions. The contribution of the paper is the improvement of the structural image recognition method with the introduction of a generalized chain description structure using the distribution values for fragments of the set of key point descriptors, which meaningfully reflect the properties of image objects and provide effective recognition. The practical significance of the paper is the achievement of an increase of image relevance calculation speed, confirmation of the effectiveness of proposed modifications on sample images, obtaining of an application software models for research and implementation of classification methods in computer vision systems.Downloads
References
Szeliski R. Computer Vision: Algorithms and Applications / R. Szeliski. – London: Springer, 2010. – 979 p.
Gadetska S.V. Statistical Measures for Computation of the Image Relevance of Visual Objects in the Structural Image Classification Methods / S.V. Gadetska, V.O. Gorokhovatsky // Telecommunications and Radio Engineering. – 2018, Vol. 77 (12), pp. 1041–1053.
Han, J. Data Mining: Concepts and Techniques / J. Han, M. Kamber. – Amsterdam e.a.: Morgan Kaufmann Publishers, 2006. – 754 p.
Gorokhovatskyi V.A. Image Classification Methods in the Space of Descriptions in the Form of a Set of the Key Point Descriptors / V.A. Gorokhovatskyi // Telecommunications and Radio Engineering. – 2018, 77 (9), pp. 787-797.
Stefan Leutenegger, Margarita Chli, Roland Y. Siegwart. BRISK: Binary Robust Invariant Scalable Keypoints. – Computer Vision (ICCV), pp. 2548 – 2555, 2011.
Гороховатський В.О. Застосування апарату аналізу та оброблення бітових даних у методах класифікації зображень за множиною ключових точок / В.О. Гороховатський, К. Г. Солодченко // Системи управління, навігації та зв’язку. – 2018. – №2 (48). – C. 63–67.
Баклицкий В.К. Методы фильтрации сигналов в корреляционно-экстремальных системах навигации/ В.К. Баклицкий, А.М. Бочкарев, М.П. Мусьяков. – М.: Радио и связь, 1986. – 216 с.
Путятин Е.П. Обработка изображений в робототехнике / Е.П. Путятин, С.И. Аверин. – М.: Машиностроение, 1990. – 320 с.
R.O. Duda, P.E.Hart and D.G. Stork, “Pattern classification”, Wiley, 2008, 738 p.
Гороховатский В.А. Структурный анализ и интеллектуальная обработка данных в компьютерном зрении: монография / В.А. Гороховатский. – Х.: Компания СМИТ, 2014. – 316с.
