基于YOLO11n的非机动车辆轻量化目标检测模型
DOI:
CSTR:
作者:
作者单位:

1.武汉理工大学机电工程学院;2.湖北天威汽车有限公司

作者简介:

通讯作者:

中图分类号:

基金项目:

校企合作攻关项目([2023]58;2023DJC082;20231h0215)资助


Lightweight object detection model for non-motor vehicles based on YOLO11n
Author:
Affiliation:

1.Wuhan University of Technology;2.Hube Tianwei Automobile Co

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    针对洒水车智能喷洒中非机动车辆检测算法复杂度高、多尺度目标检测困难等问题,提出了基于YOLO11n的轻量化目标检测模型。在YOLO11n模型的基础上,提出Lightweight Multi-scale Backneck Net骨干网络(以下简称LMBNet骨干网络),将传统下采样模块替换为Lightweight Adaptive Extraction模块(以下简称LAE模块),提出多尺度卷积模块C3k2-MSCB(C3k2-Multi-scale Convolution Block),引入大核注意力机制LSKA(Large Separable Kernel Attention)改进C2PSA,减少了模型的参数量,提升模型对多尺度信息的检测能力;使用CCFM(CNN-based Cross-scale Feature Fusion)替换原模型颈部网络,进一步缩小模型大小,提高检测效率。实验结果表明,在自建的非机动车辆数据集上,相较于基准模型,改进的YOLO11n模型的的参数量、GFLOPs和模型大小分别降低了46.3%、20.3%和42.9%,而mAP@50提升了0.3个百分点,为洒水车智能喷洒系统提供了可靠的支持。

    Abstract:

    A lightweight object detection model based on improved YOLO11n is proposed to address the issues of high complexity,and difficulty in detecting multi-scale objects in the intelligent spraying of non-motor vehicle by sprinkler trucks. On the basis of the YOLO11n model, this paper proposes lightweight multi-scale backbone network named Lightweight Multi-scale Backneck Net (LMBNet), replaces the traditional down-sampling module with Lightweight Adaptive Ext module (LAE module for short), proposes a multi-scale convolution module C3k2-Multi-scale Convolution Block (C3k2-MSCB) introduces a large core attention mechanism Large Separable Kernel Attention (LSKA) to improve C2PSA, which reduces the number of parameters of the model and improves the ability of the model for multi-scale information; use CNN-based Cross-scale Feature Fusion (CCFM) to further reduce the model size and improve detection efficiency. The experimental results showed that on the self built non-motor vehicle dataset, the improved YOLO11n model reduced the parameter count, GFLOPs, and model size by 46.3%, 20.3%, and 42.9%, respectively, compared to the baseline model mAP@50 Improved by 0.3 percentage points, providing precise and reliable support for the intelligent spraying system of sprinkler trucks.

    参考文献
    相似文献
    引证文献
引用本文
分享
相关视频

文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2025-07-17
  • 最后修改日期:2025-08-14
  • 录用日期:2025-08-14
  • 在线发布日期:
  • 出版日期:
文章二维码