基于改进YOLO的无人机入侵检测方法  

UAV intrusion detection method based on improved YOLO

作　　者：郝晋渊 张家明 张少康 张照彦 郝真鸣[2,4] 戴少石 冉宁 Hao Jinyuan;Zhang Jiaming;Zhang Shaokang;Zhang Zhaoyan;Hao Zhenming;Dai Shaoshi;Ran Ning(HBU-UCLAN School of Media,Communication and Creative Industries,Hebei University,Baoding 071002,China;College of Electronic&Informational Engineering,Hebei University,Baoding 071002,China;School of Cyber Security and Computer,Hebei University,Baoding 071002,China;Laboratory of Energy-Saving Technology,Hebei University,Baoding 071002,China;Huaneng Shang′an Power Plant,Shijiazhuang 050399,China;Laboratory of IoT Technology,Hebei University,Baoding 071002,China)

机构地区：[1]河北大学中央兰开夏传媒与创意学院,保定071002 [2]河北大学电子信息工程学院,保定071002 [3]河北大学网络空间安全与计算机学院,保定071002 [4]河北大学节能技术研发中心,保定071002 [5]华能上安电厂,石家庄050399 [6]河北大学物联网智能技术研究中心,保定071002

出　　处：《电子测量与仪器学报》2024年第7期143-151,共9页Journal of Electronic Measurement and Instrumentation

基　　金：国家自然科学基金(62373132);中央引导地方科技发展资金项目(236Z1602G);教育部“春晖计划”合作科研项目(HZKY20220257);石家庄市驻冀高校基础研究项目(241791367A);河北大学优秀青年科研创新团队建设项目(QNTD202411)资助。

摘　　要：针对现有的基于深度学习的目标检测方法在面对现实场景的无人机目标时,存在鲁棒性差、准确率低、模型复杂度高的问题,提出一种基于动态卷积的YOLO目标检测方法——OD-YOLO。该算法针对无人机目标低、慢、小的特点,采取了以下改进措施:首先针对下采样过程可能导致学习信息丢失和目标信息不突出的问题,提出空间到深度卷积来实现下采样过程,不丢失学习信息的同时突出无人机目标的特征;其次为了进一步提高目标检测的精度和对不同背景的泛化性,采用全维度动态卷积进一步提高目标检测的精度和对不同背景的泛化性;最后对模型骨干网络进行改进,增强无人机目标的语义特征,并缩减骨架大小,减少参数量,既提高模型的计算效率,又保持对无人机目标的有效表示能力。通过实验仿真,对比了OD-YOLO和当前先进的目标检测算法。结果表明,OD-YOLO在精度和轻量化方面都有显著提升。mAP和Recall分别相比原模型提高了3.4%和5.1%。In response to the limitations of existing deep learning-based object detection methods when faced with real-world unmanned aerial vehicle(UAV)targets,such as poor robustness,low accuracy,and high model complexity,a YOLO-based object detection method called OD-YOLO is proposed.This algorithm addresses the characteristics of UAV targets being small,slow,and low.Several improvements have been implemented.Firstly,to tackle the issue of learning information loss and insufficient emphasis on target information during the downsampling process,a spatial-to-depth convolution is introduced to ensure the preservation of learning information while highlighting the features of UAV targets.Secondly,to further enhance the accuracy of object detection and improve its generalization across different backgrounds,a full-dimensional dynamic convolution is be used.This enhances the accuracy of object detection and improves its generalization capabilities across various backgrounds.Lastly,the backbone network of the model is modified to enhance the semantic features of UAV targets and reduce the size of the skeleton,resulting in a reduced parameter count and improved computational efficiency of the model,while maintaining effective representation capabilities for UAV targets.Experimental simulations were conducted to compare OD-YOLO with current state-of-the-art object detection algorithms.The results demonstrate significant improvements in accuracy and lightweight performance for OD-YOLO.The mAP and Recall distributions increased by 3.4%and 5.1%,respectively,compared to the original model.

关 键 词：目标检测 无人机 双目视觉 深度学习 入侵检测 

分 类 号：TP11[自动化与计算机技术—控制理论与控制工程]