基于无人机的轻量化小目标检测网络

丛玉华; 何啸; 邢长达; 成旭明; 唐鑫; 王志胜; 欧阳权

doi:10.15892/j.cnki.djzdxb.2022.06.002

弹箭与制导学报 >

2022 , Vol. 42 >Issue 6: 6 - 12

DOI: https://doi.org/10.15892/j.cnki.djzdxb.2022.06.002

基于无人机的轻量化小目标检测网络

丛玉华 ¹^,² ,
何啸 ² ,
邢长达 ¹^,³ ,
成旭明 ¹ ,
唐鑫 ¹ ,
王志胜 ¹ ,
欧阳权 ¹

展开

¹ 南京航空航天大学自动化学院,南京 210006
² 南京理工大学紫金学院,南京 210023
³ 南京航空航天大学深圳研究院,广东深圳 518063

丛玉华(1981—),女,山东烟台人,讲师,博士研究生,研究方向:无人机集群规划与控制。

收稿日期: 2022-06-23

网络出版日期: 2025-02-25

基金资助

国家自然科学基金(62101247)

中国博士后科学基金(2022T150320)

深圳市中央引导地方科技发展专项资金(2021Szvup063)

江苏高校哲学社会科学研究项目(2021SJA2.250)

南京理工大学紫金学院校级科研项目(2022ZRKX0401004)

收起

Lightweight Small Target Detection Network Based on UAV

CONG Yuhua ¹^,² ,
HE Xiao ² ,
XING Changda ¹^,³ ,
CHENG Xuming ¹ ,
TANG Xin ¹ ,
WANG Zhisheng ¹ ,
OUYANG Quan ¹

Expand

¹ College of Automation Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing 210006, China
² Zijin College Nanjing University of Science and Technology, Nanjing 210023, China
³ Shenzhen Research Institute of Nanjing University of Aeronautics and Astronautics, Shenzhen 518063, Guangdong, China

Received date: 2022-06-23

Online published: 2025-02-25

Fold

摘要

在无人机端进行目标检测时,存在机载嵌入式设备算力有限、目标较小、背景复杂、图像分辨率低的问题。为此,目标检测网络采用YOLO体系框架进行轻量化和提升精度的改进。主干特征提取网络部分融合SPP模块,特征加强网络部分引入注意力机制,特征输出部分采用解耦头结构。通过对比测试验证了改进的算法具备实时性和高精度,适合无人机端的应用。

关键词： 无人机; YOLO; 轻量化; 注意力机制; 小目标检测

本文引用格式

丛玉华 , 何啸 , 邢长达 , 成旭明 , 唐鑫 , 王志胜 , 欧阳权 . 基于无人机的轻量化小目标检测网络[J]. 弹箭与制导学报, 2022 , 42(6) : 6 -12 . DOI: 10.15892/j.cnki.djzdxb.2022.06.002

Abstract

When target detection is carried out on the UAV terminal, there are some problems such as limited computing power of the airborne embedded equipment, small target, complex background and low image resolution. Therefore, the target detection network adopts the YOLO system framework, and on this basis, the target detection network is improved in lightweight and accuracy.The SPP module is introduced into the backbone feature extraction network part, the attention mechanism is introduced into the feature enhancement network part, and the decoupling head mechanism is used in the feature output part. The comparison test shows that the improved algorithm has real-time and high accuracy, and is suitable for the application of unmanned aerial vehicles.

Key words： UAV; YOLO; lightweight; attentional mechanism; small target detection

[an error occurred while processing this directive]

参考文献

原文顺序 | 文献年度倒序 | 文中引用次数倒序

[1]	REDMON J, FARHADI A. YOLOv3: An incremental improvement[D]. Washington: Washington University, 2018.

[2]	陈灏然. 基于卷积神经网络的小目标检测算法研究[D]. 无锡: 江南大学, 2021.

[3]	石鑫鑫. 改进的卷积神经网络算法在农作物病害检测中的应用[D]. 长春: 长春工业大学, 2021.

[4]	BOCHKOVSKIY A, WANG C Y, LIAO H Y M. YOLOv4: Optimal speed and accuracy of object detection[EB/OL].(2004-10-23)[2022-08-17]. http://https://arxiv.org/abs/2004.10934.

[5]	WANG C Y, LIAO H Y M, WU Y H, et al. CSPNet: a new backbone that can enhance learning capability of CNN[C]// IEEE. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops. New York: IEEE, 2020: 390-391.

[6]	HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition[C]// IEEE. Proceedings of the IEEE conference on computer vision and pattern recognition. New York: IEEE, 2016: 770-778.

[7]	HE K, ZHANG X, REN S, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(9): 1904-1916. DOI PMID

[8]	TAN M, PANG R, LE Q V. Efficientdet: Scalable and efficient object detection[C]// IEEE. Proceedings of Computer Vision (ECCV). New York: IEEE, 2018: 3-19.

[9]	LIU S, QI L, QIN H, et al. Path aggregation network or instance segmentation[C]// IEEE.Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2018: 8759-8768.

[10]	WOO S, PARK J, LEE J Y, et al. Cbam: Convolutional block attention module[C]// IEEE. Proceedingsof the European conference on learning for image recognition. New York: IEEE, 2016:770-778.

[11]	WANG H, WANG J, BAI K, et al. Centered multi-task generative adversarial network for small object detection[J]. Sensors, 2021, 21(15): 5194.

[12]	ZHU P, WEN L, DU D, et al. Detection and tracking meet drones challenge[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2021 (1): 11-17.

Options

文章导航