基于L-ConvNeXt网络的变电站人员安全操作检测方法

doi:10.16088/j.issn.1001-6600.2023030401

广西师范大学学报（自然科学版） ›› 2024, Vol. 42 ›› Issue (1): 102-110.doi: 10.16088/j.issn.1001-6600.2023030401

基于L-ConvNeXt网络的变电站人员安全操作检测方法

曾亮^1,2*, 胡谦^1,2, 杨腾飞^1,2, 谭微微^1,2

1.湖北工业大学电气与电子工程学院, 湖北武汉 430068;
2.太阳能高效利用及储能运行控制湖北省重点实验室(湖北工业大学), 湖北武汉 430068

收稿日期:2023-03-04 修回日期:2023-07-26 出版日期:2024-01-25 发布日期:2024-01-19
通讯作者: 曾亮(1980—), 男, 湖南临湘人, 湖北工业大学教授, 博士。E-mail:zengliang@hbut.edu.cn
基金资助:
湖北省重点研发计划项目(2020BAB114);湖北省教育厅科学研究计划重点项目(D20211402);襄阳湖北工业大学产业研究院项目(XYYJ2022C04)

Substation Personnel Safety Operation Detection Based on L-ConvNeXt Network

ZENG Liang^1,2*, HU Qian^1,2, YANG Tengfei^1,2, TAN Weiwei^1,2

1. School of Electrical and Electronic Engineering, Hubei University of Technology, Wuhan Hubei 430068, China;
2. Hubei Key Laboratory for High-efficiency Utilization of Solar Energy and Operation Control of Energy Storage System (Hubei University of Technology), Wuhan Hubei 430068, China

Received:2023-03-04 Revised:2023-07-26 Online:2024-01-25 Published:2024-01-19

摘要/Abstract

摘要： 针对变电站复杂环境中作业人员操作检测网络参数量过大、作业人员部分特征不明显等问题,本文提出一种基于L-ConvNeXt网络的变电站人员检测方法。首先,网络的主干特征提取部分由轻量化ConvNeXt模块搭建而成,保证网络特征提取能力同时使主干部分保持较低的参数量;其次,选择TPH(Transformer prediction head)作为网络末端检测头,加强网络对低分辨率特征的检测;最后,引用VariFocal Loss作为目标损失函数中的分类损失和置信度损失,进一步增加网络对正样本的损失权重。在天池公共数据集上的实验结果表明:本文网络模型获得较好的检测效果,其平均检测精度达到89.6%,模型参数量为13.2×10⁶,能够有效地检测变电站人员的作业情况,满足变电站复杂场景下的检测需求。

关键词: 目标检测, ConvNeXt, 轻量化模型, 变电站, VariFocal Loss

Abstract: A substation personnel detection based on the L-ConvNeXt network is proposed to solve the problems of excessive amount of network parameters and unclear characteristics of operational personnel detection in the complex environment of substations. Firstly, the backbone feature extraction network is built by the lightweight ConvNeXt to enhance the feature extraction capability of the network while keeping the backbone with a low parameter. Then, the transformer prediction head(TPH) is selected as the end detection head to enhance the detection of low-resolution features by the network. Finally, VariFocal Loss is introduced to replace Focal Loss as target loss and confidence loss to improve the loss weight of the network for positive samples. The experimental results on the public dataset of Tianchi show that the proposed network achieves a better detection results, the average detection accuracy reaches 89.6% and the number of model parameters is 13.2×10⁶, which can effectively detect the operation of substation personnel and meet the detection requirements of substation operators in complex scenarios.

Key words: object detection, ConvNeXt, lightweight model, substation, VariFocal Loss

中图分类号: TP391.41

曾亮, 胡谦, 杨腾飞, 谭微微. 基于L-ConvNeXt网络的变电站人员安全操作检测方法[J]. 广西师范大学学报（自然科学版）, 2024, 42(1): 102-110.

ZENG Liang, HU Qian, YANG Tengfei, TAN Weiwei. Substation Personnel Safety Operation Detection Based on L-ConvNeXt Network[J]. Journal of Guangxi Normal University(Natural Science Edition), 2024, 42(1): 102-110.

参考文献

[1] STEMN E, BOFINGER C, CLIFF D, et al. Failure to learn from safety incidents: status, challenges and opportunities[J]. Safety Science, 2018, 101: 313-325. DOI: 10.1016/j.ssci.2017.09.018.
[2] 国家能源局. 2021年一季度事故分析报告[EB/OL]. (2021-08-10)[2023-03-04]. http://www.nea.gov.cn/2021-08/10/c_1310119210.htm.
[3] 国家能源局. 2021年二季度事故分析报告[EB/OL]. (2021-09-08)[2023-03-04]. http://www.nea.gov.cn/2021-09/08/c_1310175817.htm.
[4] 杜思远. 变电站人员安全帽佩戴识别算法研究[D]. 重庆: 重庆大学, 2017.
[5] FELZENSZWALB P F, GIRSHICK R B, MCALLESTER D, et al. Object detection with discriminatively trained part-based models[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2010, 32(9): 1627-1645. DOI: 10.1109/TPAMI.2009.167.
[6] 胡金磊, 周俊煌, 林孝斌, 等. 基于S-HOG+C算子的变电作业人员着装分析方法研究[J]. 机电工程技术, 2018, 47(12): 136-140. DOI: 10.3969/j.issn.1009-9492.2018.12.043.
[7] 吴育武. 适应变电站智能安全监控的运动目标检测及人脸快速识别方法[J]. 机电工程技术, 2019, 48(11): 52-55, 126. DOI: 10.3969/j.issn.1009-9492.2019.11.019.
[8] 严良平, 童静, 王凯, 等. 基于FAST特征的智能安全帽静态报警方法的研究[J]. 电力信息与通信技术, 2019, 17(4): 67-71. DOI: 10.16543/j.2095-641x.electric.power.ict.2019.04.011.
[9] 魏明军, 周太宇, 纪占林, 等. 基于YOLOv3的公共场所口罩佩戴检测方法[J]. 广西师范大学学报(自然科学版), 2023, 41(1): 76-86. DOI: 10.16088/j.issn.1001-6600.2022030402.
[10] 李洋, 苟刚. 基于改进YOLOX 的轻量型垃圾分类检测方法[J]. 广西师范大学学报(自然科学版), 2023, 41(3): 80-90. DOI: 10.16088/j.issn.1001-6600.2022100804.
[11] FANG Q, LI H, LUO X C, et al. Detecting non-hardhat-use by a deep learning method from far-field surveillance videos[J]. Automation in Construction, 2018, 85: 1-9. DOI: 10.1016/j.autcon.2017.09.018.
[12] 杨永波, 李栋. 改进YOLOv5的轻量级安全帽佩戴检测算法[J]. 计算机工程与应用, 2022, 58(9): 201-207. DOI: 10.3778/j.issn.1002-8331.2111-0346.
[13] 丘浩, 张炜, 彭博雅, 等. 基于YOLOv3的特定电力作业场景下的违规操作识别算法[J]. 电力科学与技术学报, 2021, 36(3): 195-202. DOI: 10.19781/j.issn.1673-9140.2021.03.024.
[14] REDMON J, FARHADI A. YOLOv3: an incremental improvement[EB/OL]. (2018-04-08)[2023-03-04]. https://arxiv. org/abs/1804.02767. DOI: 10.48550/arXiv.1804.02767.
[15] 王时威. 基于深度学习的电力作业异常行为识别系统的设计与实现[D]. 武汉: 武汉纺织大学, 2021. DOI: 10.27698/d.cnki.gwhxj.2021.000120.
[16] 孙倩, 赵李强, 高雪林. 一种电力施工安全规范模型验证方法[J]. 云南电力技术, 2022, 50(3): 9-14. DOI: 10.3969/j.issn.1006-7345.2022.03.003.
[17] 刘泽西, 张楠, 连婷, 等. 针对轻量化网络的安全帽检测方法[J]. 测控技术, 2022, 41(8): 16-21, 53. DOI: 10.19708/j.ckjs.2022.08.003.
[18] 田枫, 贾昊鹏, 刘芳. 改进YOLOv5的油田作业现场安全着装小目标检测[J]. 计算机系统应用, 2022, 31(3): 159-168. DOI: 10.15888/j.cnki.csa.008359.
[19] 郭奕裕, 周箩鱼. 安全帽佩戴检测网络模型的轻量化设计[J]. 计算机工程, 2023, 49(4): 312-320. DOI: 10.19678/j.issn.1000-3428.0064219.
[20] LIU Z, MAO H Z, WU C Y, et al. A ConvNet for the 2020s[C]// 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Los Alamitos, CA: IEEE Computer Society, 2022: 11966-11976. DOI: 10.1109/CVPR52688.2022.01167.
[21] KARPOV P, GODIN G, TETKO I V. A transformer model for retrosynthesis[C]// Artificial Neural Networks and Machine Learning-ICANN 2019: Workshop and Special Sessions. Cham: Springer Nature Switzerland AG, 2019: 817-830. DOI: 10.1007/978-3-030-30493-5_78.
[22] HUANG J H, FANG Y Y, WU Y Z, et al. Swin transformer for fast MRI[J]. Neurocomputing, 2022, 493: 281-304. DOI: 10.1016/j.neucom.2022.04.051.
[23] LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(2): 318-327. DOI: 10.1109/TPAMI.2018.2858826.
[24] ZHANG H Y, WANG Y, DAYOUB F, et al. VarifocalNet: an IoU-aware dense object detector[C]// 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Los Alamitos, CA: IEEE Computer Society, 2021: 8510-8519. DOI: 10.1109/CVPR46437.2021.00841.
[25] MA N N, ZHANG X Y, ZHENG H T, et al. ShuffleNet V2: practical guidelines for efficient CNN architecture design[C]// Computer Vision-ECCV 2018: LNCS Volume 11218. Cham: Springer Nature Switzerland AG, 2018: 122-138. DOI: 10.1007/978-3-030-01264-9_8.
[26] ZHAO G S, WANG Y, WANG J. Lightweight intrusion detection model of the Internet of things with hybrid cloud-fog computing[J]. Security and Communication Networks, 2023, 2023: 7107663. DOI: 10.1155/2023/7107663.
[27] 王媛彬, 李媛媛, 段誉, 等. 基于轻量骨干网络和注意力结构的变电设备红外图像识别[J]. 电网技术, 2023, 47(10): 4358-4369.DOI: 10.13335/j.1000-3673.pst.2022.2113.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

基于L-ConvNeXt网络的变电站人员安全操作检测方法

Substation Personnel Safety Operation Detection Based on L-ConvNeXt Network

PDF (PC)

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 4

Metrics

本文评价

推荐阅读 10

[1]	黄叶祺, 王明伟, 闫瑞, 雷涛. 基于改进的YOLOv5金刚石线表面质量检测[J]. 广西师范大学学报（自然科学版）, 2023, 41(4): 123-134.
[2]	陈文康, 陆声链, 刘冰浩, 李帼, 刘晓宇, 陈明. 基于改进YOLOv4的果园柑橘检测方法研究[J]. 广西师范大学学报（自然科学版）, 2021, 39(5): 134-146.
[3]	刘士李, 朱晓虎, 刘丽, 方天睿. 基于模糊平滑的安徽地区110 kV GIS 变电站LCC 模糊估算模型[J]. 广西师范大学学报（自然科学版）, 2020, 38(5): 24-33.
[4]	刘英璇, 伍锡如, 雪刚刚. 基于深度学习的道路交通标志多目标实时检测[J]. 广西师范大学学报（自然科学版）, 2020, 38(2): 96-106.