Article 4318

Title of the article



Borisova Irina Valentinovna, Candidate of engineering sciences, associate professor, sub-department of autonomous information and control systems, Novosibirsk State Technical University (20 K. Marxa avenue, Novosibirsk, Russia), E-mail:
Legkiy Vladimir Nikolaevich, Doctor of engineering sciences, associate professor, head of sub-department of autonomous information and control systems, Novosibirsk State Technical University (20 K. Marxa avenue, Novosibirsk, Russia), E-mail:
Utev Dmitriy Andreevich, Student, Novosibirsk State Technical University (20 K. Marxa avenue, Novosibirsk, Russia), E-mail:
Kravets Sergey Aleksandrovich, Postgraduate student, Novosibirsk State Technical University (20 K. Marxa avenue, Novosibirsk, Russia), E-mail:
Demidov Dmitriy Evgen'evich, Postgraduate student, Novosibirsk State Technical University (20 K. Marxa avenue, Novosibirsk, Russia), E-mail: 

Index UDK





Background. The analysis of proximity measures for comparison of images from different spectral ranges is carried out for the problem of search targets corresponding to the standard. Fragments of real images and generated pseudo-images are used as standards. The goal of the research is to choose a measure of proximity that provides stable object detection by their reference image.
Materials and methods. Object detection is performed by searching for the extremum of the criterion function. The following proximity measures are considered: correlation, comparison, and Chamfer Distance. Analysis of proximity measures is carried out in television and thermal images of the simulated scene with models of ground equipment and complex background.
Results. It is shown that the contour and gradient orientation provide fewer errors of target detection compared to brightness. The width of the contour in binary images must be consistent with the number and size of the background elements.
Conclusions. The percentage of correct detection is higher when using a comparison measure than using normalized and binary correlation. The best results were obtained using Chamfer Distance, which does not require an exact match of the contours. This is important if the object and the standard don’t match perfectly. 

Key words

image processing, target detection, measure of proximity, thermal image, gradient orientation 

 Download PDF

1. Popov P. G., Borisova I. V. Opticheskiy zhurnal [Optical journal]. 1999, vol. 66, no. 4, pp. 94–101.
2. Blokhinov Yu. B., Chernyavskiy A. S. Mekhanika, upravlenie i avtomatika [Mechanics, control and automatics]. 2012, no. 8, pp. 181–188.
3. Tsvetkov O. V., Tananykina L. V. Komp'yuternaya optika [Computer optics]. 2015, vol. 39, no. 5, pp. 738–743.
4. Novikov A. I., Sablina V. A., Goryachev E. O. Izvestiya Tul'skogo gosudarstvennogo universiteta. Tekhnicheskie nauki [Proceedings of Tula State University. Engineering sciences]. 2013, iss. 9, part 1, pp. 260–270.
5. Zitova B., Flusser J. Image and Vision Computing. 2003, vol. 21, pp. 977–1000.
6. Borgefors G. Computer Vision, Graphics, and Image Processing. 1986, vol. 34, no. 3, pp. 344–371.
7. Borisova I. V., Legkiy V. N., Kravets S. A. Komp'yuternaya optika [Computer optics]. 2017, vol. 41, no. 6, pp. 931–937.
8. Otsu N. IEEE Transactions on systems, MAN, and CYBERNETICS. 1979, vol. SMC-9, no. 1, pp. 62–66.


Дата создания: 19.04.2019 14:01
Дата обновления: 22.04.2019 08:08