Optical multi-band detection of unmanned aerial vehicles with YOLO v4 convolutional neural network

被引：3

作者：

Golyak, Igor S. ^{[1
]}

Anfimov, Dmitry R. ^{[1
]}

Fufurin, Igor L. ^{[1
]}

Nazolin, Andrey L. ^{[1
]}

Bashkin, Sergey, V ^{[1
]}

Glushkov, Vladimir L. ^{[1
]}

Morozov, Andrey N. ^{[1
]}

机构：

[1] Bauman Moscow State Tech Univ, Moscow 105005, Russia

来源：

SPIE FUTURE SENSING TECHNOLOGIES (2020) | 2020年 / 11525卷

关键词：

unmanned aerial vehicle; drone detection; convolutional neural network; safety; remote sensing;

D O I：

10.1117/12.2584591

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

The paper presents a method for optical detection drones using the YOLO v4 neural network. Recognition performs simultaneously in the visible (Vis) and long-wave infrared (LWIR) ranges. The results of UAV detection on various types of urban background environment at day and night conditions, as well as at different distances from cameras, are presented. An algorithm for detecting of unmanned vehicles in the video cameras field of view of the Vis and LWIR ranges is described. This algorithm takes as input the outputs of two neural networks that recognize the drone in two ranges and estimates the probability of detection. Its shown that the YOLO v4 neural network recognizes unmanned objects on various background substrates with a minimum temperature difference of 0.4 degrees on the NEC 2640 thermal imager.

引用

页数：6