智能科学与技术学报 ›› 2021, Vol. 3 ›› Issue (3): 312-321.doi: 10.11959/j.issn.2096-6652.202132

• 专刊:目标智能检测与识别 • 上一篇    

基于SE-YOLOv5s的绝缘子检测

田庆1, 胡蓉1, 李佐勇2,3, 蔡远征2,3, 余兆钗2,3   

  1. 1 福建工程学院计算机科学与数学学院,福建 福州 350118
    2 闽江学院计算机与控制工程学院,福建 福州 350121
    3 福建省信息处理与智能控制重点实验室,福建 福州 350121
  • 修回日期:2021-07-27 出版日期:2021-09-01 发布日期:2021-09-01
  • 作者简介:田庆(1998− ),男,福建工程学院计算机科学与数学学院硕士生,主要研究方向为深度学习、计算机视觉
    胡蓉(1974− ),女,博士,福建工程学院计算机科学与数学学院教授,主要研究方向为人工智能、大数据、交通数据挖掘等
    李佐勇(1980− ),男,博士,闽江学院计算机与控制工程学院教授,主要研究方向为计算机视觉、模式识别、深度学习等
    蔡远征(1987− ),男,博士,闽江学院计算机与控制工程学院讲师,主要研究方向为计算机视觉与模式识别
    余兆钗(1979− ),男,闽江学院计算机与控制工程学院讲师,主要研究方向为医学图像分析与模式识别
  • 基金资助:
    福建省自然科学基金资助项目(2020J02024);福建省自然科学基金资助项目(2020J01828);福州市科技计划资助项目(2020-RC-186)

Insulator detection based on SE-YOLOv5s

Qing TIAN1, Rong HU1, Zuoyong LI2,3, Yuanzheng CAI2,3, Zhaochai YU2,3   

  1. 1 College of Computer Science and Mathematics, Fujian University of Technology, Fuzhou 350118, China
    2 College of Computer and Control Engineering, Minjiang University, Fuzhou 350121, China
    3 Fujian Provincial Key Laboratory of Information Processing and Intelligent Control, Fuzhou 350121, China
  • Revised:2021-07-27 Online:2021-09-01 Published:2021-09-01
  • Supported by:
    The Natural Science Foundation of Fujian Province(2020J02024);The Natural Science Foundation of Fujian Province(2020J01828);Fuzhou Science and Technology Project(2020-RC-186)

摘要:

在电力系统需要巡检的大环境下,人工巡检的传统方式存在很大不便和安全隐患,而采用无人机的目标检测方法在绝缘子检测识别方向有很大的应用前景。针对绝缘子图像检测中存在的场景复杂、视角多变、设备计算能力受限等问题,提出了一种改进的轻量级SE-YOLOv5s卷积神经网络来实现对绝缘子的快速目标检测,该方法首先在YOLOv5s网络中融合SE注意力模块,以强化网络对绝缘子目标的辨识能力,随后采用K-means聚类方法构建绝缘子的先验框,以提升定位精度,最后构造置信度与定位任务联合的损失函数,并引入 Mosaic 数据增强策略训练网络,有效解决训练数据不足的问题。经过实验验证发现,与主流目标检测方法相比,提出的SE-YOLOv5s方法在绝缘子检测准确率、召回率、检测速度及平均精度均值等性能指标上均取得了较好的结果。实验结果表明,该网络对于绝缘子检测有很好的效果,具有更好的鲁棒性,对电力系统的巡检方式具有参考价值。

关键词: 绝缘子, 目标检测, 深度学习, YOLOv5s

Abstract:

In a large environment where the power system needs to be inspected, the traditional method of manual inspection has great inconvenience and potential safety risks, and the object detection method of unmanned aerial vehicle has great application prospects in the direction of insulator detection and recognition.SE-YOLOv5s, a lightweight insulator detection network that performs efficient detection for this task was presented.Firstly, backbone of YOLOv5s by fusing the SE attention module was strengthen.Then, the position distribution of insulator object was investigated and predefined templates of a priori box by K-means clustering on prior coordinate vectors were generated.Finally, the network by a multitask loss function was trained combined with confidence and position regression task.Furthermore, Mosaic data augmentation was utilized to supplement additional training samples.Experimental results demonstrate that the proposed SE-YOLOv5s significantly outperforms baseline methods at multiple criterions including accuracy, recall rate, detection rate and mean average precision.In comparison with the baselines, the proposed network has a flexible trade-off between robustness and memory overhead and it is a potential approach to promote the power system development.

Key words: insulator, object detection, deep learning, YOLOv5s

中图分类号: 

No Suggested Reading articles found!