Issue |
JNWPU
Volume 36, Number 2, April 2018
|
|
---|---|---|
Page(s) | 258 - 263 | |
DOI | https://doi.org/10.1051/jnwpu/20183620258 | |
Published online | 03 July 2018 |
Small UAV Target Detection Model Based on Deep Neural Network
基于深度神经网络的低空弱小无人机目标检测研究
1
National Key Laboratory of Aerospace Flight Dynamics, Xi'an 710072, China
2
School of Astronautics, Northwestern Polytechnical University, Xi'an 710072, China
Received:
12
June
2017
Unmanned aerial vehicle (UAV) has relatively small size and weak visual characteristics. The recognition accuracy of traditional object detection methods can decrease sharply when complex background and distraction objects exist. In this paper, we proposed a novel deep neural network (DNN) model for small UAV target recognition task. Based on the visual characteristics of surveillance image and UAV target, a multi-channel DNN is designed. Training and optimization of the DNN are completed with self-constructed UAV image database. Simulation results show that the proposed DNN model can achieve good results in recognizing the variable-scale UAV target and have compatible performance in distinguishing the interference and that the proposed model is robust and has a great potential prospect for engineering application.
摘要
针对低空无人机目标视觉特征较弱,传统识别模型在目标尺度较小时易受干扰导致识别精度下降等问题,提出了一种基于多隐含层深度神经网络的弱小无人机目标检测模型。根据低空监视图像输入特性和弱小无人机目标视觉表征特点,设计了包含多个隐含层的多通道深度神经网络模型结构,并通过建立多尺度、多角度、多背景条件下的无人机目标图像数据库,完成了对深度网络模型参数的训练及优化。仿真结果表明,所设计的深度模型对低空无人机目标具有较好的变尺度检测能力和抗干扰效果,体现出良好的鲁棒性和潜在工程应用前景。
Key words: unmanned aerial vehicle(UAV) / object recognition / deep neural network(DNN) / multi-hidden layer / neural networks / optimization
关键字 : 低空无人机 / 目标识别 / 深度神经网络 / 多隐含层
© 2018 Journal of Northwestern Polytechnical University. All rights reserved.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.