The comparison of pattern recognition algorithms for semantic image analysis

Authors: Vasileva E.A.
Published in issue: #3(32)/2019
DOI: 10.18698/2541-8009-2019-3-447
Category: Mechanical Engineering and Machine Science | Chapter: Robots, Mechatronics, and Robotic Systems
Keywords: computer vision, pattern recognition, machine learning, image classification, neural network, SVM algorithm, CatBoost algorithm, LightGBM algorithm, SGDClassifier algorithm
Published: 04.03.2019

The work is devoted to the study of the problem of the agricultural sector autonomous control within the scope of limited information about the surrounding area. The problem of detecting depressed areas with the help of unmanned aerial vehicles equipped with a computer vision system is considered. The modern algorithms for the object classification have been presented; a description of pattern recognition algorithms has been made. A comparative analysis of image recognition algorithms based on linear methods and methods using gradient boosting has been performed. The results of testing the algorithms performance are presented. As a result of the analysis, it was found out that the SGDClassifier algorithm showed the best recognition quality based on the Precision, Recall and F-measure metrics.

References

[1] Kalyaev I.A., Gayduk A.R., Kapustyan S.G. Modeli i algoritmy kollektivnogo upravleniya v gruppakh robotov [Collective direction models and algorithms in robot groups]. Moscow, Fizmatlit Publ., 2009 (in Russ.).

[2] Rubtsov I.V., Mashkov K.Yu., Naumov V.N. [185-anniversary of Bauman MSTU: current state and future prospects of special robotics development]. Sb. mat. 10 Vseros. nauch.-prakt. konf. Perspektivnye sistemy i zadachi upravleniya [Proc. 10th Russ. Sci.-Pract. Conf. Future systems and control problems], 2015, no. 1, pp. 247–255 (in Russ.).

[3] Vasil’yeva E.A. Developing the algorithm of the machine vision system for the images semantic analysis. Politekhnicheskiy molodezhnyy zhurnal [Politekhnicheskiy molodezhnyy zhurnal [Politechnical student journal], 2018, no. 9(26). DOI: 10.18698/2541-8009-2018-9-366 URL: http://ptsj.ru/catalog/menms/robots/366.html (in Russ.).

[4] Mestetskiy L.M. Matematicheskie metody raspoznavaniya obrazov [Mathematical methods of image recognition]. Moscow, MSU Publ., 2002–2004 (in Russ.).

[5] Chen L.-C., Papandreou G., Kokkinos I., et al. DeepLab: semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected CRFs. IEEE Trans. Pattern Anal. Mach. Intell., 2018, vol. 40, no. 4, pp. 834–848. DOI: 10.1109/TPAMI.2017.2699184 URL: https://ieeexplore.ieee.org/document/7913730

[6] Cristianini N., Shawe-Taylor J. An introduction to support vector machines and other kernel-based learning methods. Cambridge University Press, 2000.

[7] Prokhorenkova L., Gusev G., Vorobev A., et al. CatBoost: unbiased boosting with categorical features. arxiv.org: website. URL: https://arxiv.org/abs/1706.09516 (accessed: 01.12.2018).

[8] Ke G., Meng Q., Finley T., et al. LightGBM: a higly efficient gradient boosting decision tree. Proc. NIPS 2017. URL: https://papers.nips.cc/paper/6907-lightgbm-a-highly-efficient-gradient-boosting-decision-tree (accessed: 01.12.2018).

[9] Zhang T. Solving large scale linear prediction problems using stochastic gradient descent algorithms. Proc. ICML, 2004, p. 116. DOI: 10.1145/1015330.1015332 URL: https://dl.acm.org/citation.cfm?id=1015332