ROMJIST Volume 23, No. S, 2020, pp. S53-S66
Dan-Marius DOBREA, Monica-Claudia DOBREA An autonomous UAV system for video monitoring of the quarantine zones
ABSTRACT: In this paper, a comparative study between two classification approaches was done, namely, between a Support Vector Machine (SVM) neural network - as a classical machine learning approach -, and four different deep learning classification systems, considered today to be the state-of-the-art systems in computer vision. Two embedded companion computers, a Raspberry Pi and a Jetson Nano, placed on a HoverGames quadcopter support the classification systems that were developed and deployed to identify humans. The ultimate goal of this research consists in developing a quadcopter system endowed with the capabilities of following a pre-programmed flight route and simultaneously detecting humans as well as of warning the system operator to reinforce the quarantine zones. The obtained results demonstrated the superior performances provided by the deep learning approach: more than six times faster than the classical approach, and with a correct classification performance higher than 90% on a direct stream of video data. KEYWORDS: UAV, HoverGames, PX4, QGroundControl, deep learning, HOG, SVMRead full text (pdf)