基于AC-YOLO的路面落叶检测方法
DOI:
作者:
作者单位:

1.中国矿业大学;2.上海交通大学;3.徐州易尔环保科技有限公司

作者简介:

通讯作者:

中图分类号:

TP391.4

基金项目:

国家自然科学基金项目(No.61976218),徐州市重点科技项目(No.KC19072),中央高校基本科研业务费专项资金资助(No.2020ZDPY0303),江苏省研究生科技创新项目(No.KYCX21_2262),江苏省研究生科技创新项目(No.SJCX21_0992),江苏省研究生科技创新项目(No.SJCX21_1034)


A Detection Method of Fallen Leaves on Road Based on AC-YOLO
Author:
Affiliation:

1.China University of Mining and Technology;2.Shanghai Jiao Tong University

Fund Project:

the General Program of National Natural Science Foundation of China(Grant No.61976218), the Key Technology Project of Xuzhou(Grant No.KC19072), the Fundamental Research Funds for the Central Universities(Grant No.2020ZDPY0303), Graduate Research and Innovation Projects of Jiangsu Province(Grant No.KYCX21_2262), Graduate Research and Innovation Projects of Jiangsu Province(Grant No.SJCX21_1034), Graduate Research and Innovation Projects of Jiangsu Province(Grant No.SJCX21_0992)

  • 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:

    随着目前城市绿化程度的不断提高,落叶清理任务变得更加复杂繁重。针对落叶形状多变、大小不一、背景复杂、分布不均的特点,本文提出一种融合Attention-Context(AC)网络和YOLOv3的落叶检测算法(AC-YOLO),解决现有模型对落叶漏检、误检的问题,实现快速、准确地识别检测路面落叶。针对小目标落叶易发生漏检的问题,提出了AC网络结构,将不同层次的特征映射融合作为小目标的上下文信息,同时引入自注意力机制抑制复杂背景和底层噪声带来的影响,提升小目标落叶检测能力;其次,采用Mish激活函数替换Leaky ReLU,提升模型的泛化能力,提高落叶检测准确度;最后,考虑到落叶堆叠情况对清理机器人的工作效率有影响,提出了非极大值融合算法(Non-Maximum Fusion ,NMF)融合密集落叶预测框,从而通过更少的导航点解决密集落叶的检测问题,同时提升落叶检测清理的效率。实验结果表明,基于 AC--YOLO 的检测算法对落叶检测的覆盖率 (Cover) 达到 95%,检测速度达到每秒 53 帧,可以完成实际应用环境中的落叶检测任务,实现对落叶的高效率、智能化清理。。

    Abstract:

    With the continuous improvement of urban greening, it is more complex and heavy to clean the fallen leaves. According to the characteristics of variable shape, complex background and uneven distribution of fallen leaves, we proposes a fallen leaf detection algorithm, AC-YOLO, Which integrates attention context (AC) network and YOLOv3, to detect fallen leaves on road quickly and accurately. To solve the problem that small leaves are difficult to detect, we proposed AC network. We use different feature levels as the context information of small leaves and attention mechanism to suppress the influence of complex background and bottom noise, so as to improve the ability of small leaves detection. In addition, we use the Mish activation function to replace Leaky ReLU to enhance the generalization ability of the model and improve the accuracy of leaf detection. Finally, considering that the stacking of fallen leaves has an impact on the work efficiency of the cleaning robot, we proposed the Non-Maximum Fusion to fuse the dense fallen leaves detection boxes. It promotes the detection of dense fallen leavese and improves the work efficiency by reducing goal nodes. Experiment results show that the Cover of leaf detection based on AC-YOLO algorithm is 95% and the detection speed is 53 frames per second. It can complete the leaf detection task in the practical application and realize the efficient and intelligent leaves cleaning.

    参考文献
    相似文献
    引证文献
引用本文
分享
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:2021-10-13
  • 最后修改日期:2022-03-15
  • 录用日期:2022-03-28
  • 在线发布日期:
  • 出版日期: