Investigation of efficiency of detection and recognition of drone images from video stream of stationary video camera

Зубков, О. В.; Шейко, С. О.; Олейніков, В. М.; Карташов, В. М.; Бабкін, С. І.

Публікація:
Investigation of efficiency of detection and recognition of drone images from video stream of stationary video camera

dc.contributor.author	Зубков, О. В.
dc.contributor.author	Шейко, С. О.
dc.contributor.author	Олейніков, В. М.
dc.contributor.author	Карташов, В. М.
dc.contributor.author	Бабкін, С. І.
dc.date.accessioned	2021-11-26T08:54:03Z
dc.date.available	2021-11-26T08:54:03Z
dc.date.issued	2021
dc.description.abstract	An algorithm was developed for detecting and recognizing drones in the video stream of a stationary camera, which provides a high processing speed, significant detection range, and high recognition accuracy. This algorithm consists of detection of moving objects that are classified using a neural network. To detect moving objects, the authors used methods for distinguishing moving objects against a moveless background and analyzing the history of movement. The efficiency of applying the background models of MOG, MOG2, KNN, GMG, CNT, GSOC images was analyzed. Selection of these models' parameters was formulated. High performance and low noise level were the criteria for selection. Models of fully connected and convolutional neural networks were created to classify 12 types of moving objects. Sets of images, such as drones, fragments of tree foliage, grass, clouds, and insects were created to train neural networks. Recommendations were given for the number of network layers, the number of neurons in a layer, the number of convolutions to achieve maximum performance, and recognition accuracy. Comparative analysis of the accuracy of drone classification using fully connected and convolutional networks were proven effective. The dependence of the drone detection accuracy on the image size and, accordingly, on the distance to this drone was plotted. Analysis of experimental data processing under various weather and seasonal conditions showed a high efficiency of the developed algorithm at ranges up to 60 m when using a FullHD video camera with a viewing angle of 60°. In this case the maximum detection range reached 120 m	uk_UA
dc.identifier.citation	Investigation of efficiency of detection and recognition of drone images from video stream of stationary video camera / O. V. Zubkov, S. O. Sheiko, V. M. Oleynikov, V. M. Kartashov, S. I. Babkin // Telecommunications and Radio Engineering. Volume 80, 2021 Issue 3,doi: 10.1615/TelecomRadEng.2021036535	uk_UA
dc.identifier.uri	https://openarchive.nure.ua/handle/document/18395
dc.language.iso	en	uk_UA
dc.subject	drone detection	uk_UA
dc.subject	pattern recognition	uk_UA
dc.subject	machine learning	uk_UA
dc.subject	alarm and security	uk_UA
dc.subject	convolutional network	uk_UA
dc.subject	background modeling	uk_UA
dc.subject	background subtraction	uk_UA
dc.title	Investigation of efficiency of detection and recognition of drone images from video stream of stationary video camera	uk_UA
dc.type	Conference proceedings	uk_UA
dspace.entity.type	Publication