基于改进YOLOv4的无人机航拍目标检测算法  

UAV aerial object detection algorithm based on improved YOLOv4

作　　者：赵耘彻 张文胜 刘世伟 Zhao Yunche;Zhang Wensheng;Liu Shiwei(Key Laboratory of Traffic Safety and Control of Hebei Province,Shijiazhuang 050043,China;School of Traffic and Transportation,Shijiazhuang Tiedao University,Shijiazhuang 050043,China;Cangzhou Bohai New Area Huanghua Transportation Bureau,Huanghua 061100,China)

机构地区：[1]河北省交通安全与控制重点实验室,石家庄050043 [2]石家庄铁道大学交通运输学院,石家庄050043 [3]沧州渤海新区黄骅市交通运输局,黄骅061100

出　　处：《电子测量技术》2023年第8期169-175,共7页Electronic Measurement Technology

基　　金：中央引导地方科技发展资金(226Z6101G);河北省及石家庄市引进国外智力项目(2021-2023);石家庄市科技计划项目(221130134A)资助。

摘　　要：针对无人机航拍目标检测对检测速度的高要求以及航拍图像小目标较多时易出现漏检、误检的问题,提出一种基于改进YOLOv4的无人机航拍目标检测算法。首先,引入轻量级网络MobileNetv3替换YOLOv4的主干特征提取网络,并采用深度可分离卷积替换网络其余部分的3×3标准卷积,降低了模型复杂度,提升了检测速度;其次,增加了针对小目标的104×104的浅层检测层,将原特征提取网络的3种检测尺度扩展为4种,对应加深特征融合网络层数,提升了算法对小目标的检测精度;最后,采用K-means++聚类算法重新设计了初始锚框,加快了网络的收敛速度。在无人机航拍数据集上进行了对比实验,结果表明所提算法与原算法相比,在保证平均检测精度的同时,提升了小目标检测精度,且模型参数量减少了60%,检测速度提升了15.2%,在实时性和准确性方面均有较好性能。A UAV aerial object detection algorithm based on modified YOLOv4is proposed in order to address the high need of detection speed for UAV aerial object detection as well as the issue of missed detection and false detection when there are many small targets in aerial images.Firstly,the lightweight network MobileNetv3is introduced to replace the main feature extraction network of YOLOv4,and the depth separable convolution is used to replace the 3×3standard convolution of the network,which reduces the complexity of the model and improves the detection speed.Secondly,the 104×104shallow detection layer for small targets is added,the three detection scales of the original feature extraction network are increased to four,and the number of feature fusion network layers is increased.These changes increase the algorithm′s accuracy in detecting small targets.Finally,the K-means++clustering technique is used to redesign the initial anchor frame,accelerating the network′s rate of convergence.The UAV aerial data set is used in a comparison experiment.The findings demonstrate that as compared to the original approach,the suggested technique not only ensures average detection accuracy but significantly enhances the detection accuracy of small targets.The detection time is 15.2%faster while the model parameters are lowered by 60%.It performs accurately and with good real-time performance.

关 键 词：目标检测 无人机 YOLOv4 轻量级网络 K-means++ 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] V19[自动化与计算机技术—计算机科学与技术]