ABSTRACTS OF ARTICLES OF THE JOURNAL "INFORMATION TECHNOLOGIES".
No. 2. Vol. 25. 2019

DOI: 10.17587/it.25.116-127

O. S. Amosov, D. Sc., Prof., Principal Researcher, e-mail: osa18@yandex.ru, S. G. Amosova, Ph. D., Assistant Professor, Senior Researcher, e-mail: amosovasg@yandex.ru, V. A. Trapeznikov Institute of Control Sciences of RAS, Y. S. Ivanov, Ph. D., Assistant Professor, e-mail: ivanov_ys@icloud.com, S. V. Zhiganov, Postgraduate Student, e-mail: zhiganov@knastu.ru, Komsomolsk-na-Amure State University

Modelling of Intelligent Access Monitoring and Control System for Vehicles with Using the Deep Neural Networks

Mathematical model of intelligent access monitoring and control system for vehicle is developed. It differs from the existing ones as it allows generating the control actions during the handling of normal and abnormal situations for a significant reduction of recognition errors. The realization of vehicle localization using the YOLO deep neural network, allowing additional determining the type of the access object, is proposed. The solution of the license plate localization and recognition problem is based on the composition of the traditional image processing methods and the two-pass classification carried out by the modified architecture of the MobileNet convolutional network. It has been experimentally proved that the application of the developed approach gives a percentage of correct license plates recognition in the video stream of not less than 96 %, depending on external conditions. The programs complex by using Python is realized.
Keywords: mathematical model, monitoring and control system, vehicle, localization, recognition, convolutional neural network, MobileNet

Acknowlegement: The work was financially supported by the Russian Ministry of Education and Science — the project
¹ 2.1898.2017/4.6 "Designing the Mathematical and Algorithmic Ware of Intelligent Information and Telecommunication System for Higher Educational Institution Security".

 P. 116–127

To the contents