摘要
针对无人机在复杂户外环境中进行搜救任务时小目标检测性能不足的问题,文中提出一种改进的YOLOv8小目标检测模型(YOLOv8-CFT)。该模型通过引入C2F-SWC模块增强特征提取能力,同时结合FADPN网络实现高效特征融合,进一步利用TDH模块优化检测头以平衡检测精度和效率。将YOLOv8-CFT模型在无人机搜救数据集UAVSRD上进行实验,YOLOv8-CFT模型在精确率、召回率和mAP等指标上均显著优于基准YOLOv8模型,其中mAP@0.5和mAP@0.5:0.95分别提升了3.8%和8.9%。实验结果表明,YOLOv8-CFT模型在无人机搜救任务中具有更好的小目标检测能力。
A small object detection model YOLOv8⁃CFT based on improved YOLOv8 is proposed.It aims to improve the small object detection performance when UAVs perform search and rescue missions in complex outdoor environments.This model strives to enhance feature extraction capabilities by introducing the C2F⁃SWC module.Meanwhile,it combines FADPN(feature aggregation and diffusion pyramid network)to realize efficient feature fusion.Further,the TDH module is used to optimize the detection head to balance detection accuracy and detection efficiency.The YOLOv8⁃CFT model was tested on the UAV search and rescue dataset UAVSRD.The YOLOv8⁃CFT model is significantly better than the benchmark YOLOv8 model in terms of precision rate,recall rate and mAP,among which mAP@0.5 and mAP@0.5:0.95 increase by 3.8%and 8.9%,respectively.Experimental results show that the YOLOv8⁃CFT model has better small object detection capabilities in UAV search and rescue missions.
作者
金思雨
李嘉诚
黄岚
陈中举
詹炜
JIN Siyu;LI Jiacheng;HUANG Lan;CHEN Zhongju;ZHAN Wei(College of Computer Science,Yangtze University,Jingzhou 434023,China)
出处
《现代电子技术》
北大核心
2025年第17期167-175,共9页
Modern Electronics Technique
基金
湖北省教育厅教研项目(2021268)
国家自然科学基金资助项目(62276032)
中国高校产学研创新基金:新一代信息技术创新项目(2023IT269)。
