基于RLSW-YOLOv8n的无人机航拍小目标检测模型OA
Small Object Detection Model of UAV Aerial Imagery Based on RLSW-YOLOv8n
针对无人机航拍小目标检测模型精度低和参数量大的问题,提出RLSW-YOLOv8n模型:在骨干网络中构造C2f-RG模块以提升特征提取能力;设计LCFFN-P2重构颈部网络,在减少参数量的同时实现多尺度特征融合;提出SWSimAM模块以突出小目标特征.实验结果表明,改进模型在VisDrone2019数据集上相较原模型检测精度提高了8.1%,参数量减少了32.2%,性能有所提升.
Aiming at the problems of low accuracy and large parameters in small object detection models for UAV aerial imagery,this paper proposes the RLSW-YOLOv8n model.The C2f-RG module is constructed in the backbone network to enhance feature extraction capability.LCFFN-P2 is designed to reconstruct the neck network,achieving multi-scale feature fusion while reducing parameters.The SWSimAM module is proposed to highlight the features of small objects.Experimental results on the VisDrone2019 dataset show that compared with the original model,the improved model increases detection accuracy by 8.1%and reduces parameters by 32.2%.It also outperforms several existing state-of-the-art methods.
王统;安洋;赵利辉;孟迪
中北大学软件学院,太原 030051中北大学软件学院,太原 030051中北大学软件学院,太原 030051北方自动控制技术研究所,太原 030006
信息技术与安全科学
YOLOv8n无人机小目标检测多尺度特征融合SWSimAM
YOLOv8nUAVsmall object detectionmulti-scale feature fusionSWSimAM
《火力与指挥控制》 2026 (4)
61-73,13
山西省青年科学研究项目(202203021212114)中北大学第20届研究生科技立项资助项目(20242058)
评论