首页|期刊导航|郑州大学学报(理学版)|基于改进YOLOv11n的无人机小目标检测算法

基于改进YOLOv11n的无人机小目标检测算法OA

UAV Small Target Detection Algorithm Based on Improved YOLOv11n

中文摘要英文摘要

针对无人机视角下小目标检测中存在的目标尺度多变、密集遮挡及背景干扰等问题,提出一种基于改进YOLOv11n 的小目标检测算法.首先,引入多模态特征增强注意力模块,通过融合多维度的特征信息强化模型对关键区域的感知能力,从而提高特征表达的完整性与判别性.其次,构建 MS_C3K2 模块,采用不同感受野的卷积核并行处理输入特征,有效增强了局部与全局信息的提取与融合能力.最后,为进一步提升特征融合阶段的适应性与动态性,提出了自适应门控双向特征金字塔网络,通过动态融合多尺度特征提升网络对关键信息的关注与响应能力.实验结果表明,改进模型的 mAP50 达到 33.1%,mAP50-95 达到 18.9%,相比基线模型分别提升 3.4 和 2.3 个百分点,验证了所提算法在无人机复杂场景下具有良好的检测性能与鲁棒性.

To address the challenges of target scale variation,dense occlusion,and background interfer-ence in small target detection from UAV perspectives,a detection algorithm based on improved YOLOv11n was proposed.Firstly,a multi-modal feature enhancement attention module was introduced to strengthen the model's perception of key regions through integrating multi-dimensional feature informa-tion,thereby improving the completeness and discriminability of feature representation.Secondly,an MS_C3K2 module was constructed,which utilized convolutional kernels with different receptive fields to process input features in parallel,to effectively enhance the ability to extract and fuse both local and global information.Finally,to further enhance the adaptability and dynamism of the feature fusion stage,an adaptive gated bi-directional feature pyramid network was proposed to dynamically fuse multi-scale fea-tures,to enhance the network's attention and response capability to key information.Experimental results demonstrated that the mAP50 of the improved model reached 33.1%,mAP50-95 reached 18.9%,which was 3.4 and 2.3 percentage points higher than the baseline model,respectively,validating its excellent de-tection performance and robustness in complex UAV scenarios.

姚越;杨晓弘;高志威;唐强

湖南工业大学 空天技术学院 湖南 株洲 412007湖南工业大学 空天技术学院 湖南 株洲 412007湖南工业大学 计算机与人工智能学院 湖南 株洲 412007湖南工业大学 空天技术学院 湖南 株洲 412007

信息技术与安全科学

YOLOv11小目标检测注意力机制无人机双向特征金字塔网络

YOLOv11small target detectionattention mechanismUAVbi-directional feature pyra-mid network

《郑州大学学报(理学版)》 2026 (6)

32-38,7

湖南省教育厅科学研究重点项目(23A0446)

10.13705/j.issn.1671-6841.2025110

评论