基于TR-YOLOv5的输电线路多类缺陷目标检测方法

doi:10.11996/JG.j.2095-302X.2023040667

图学学报 ›› 2023, Vol. 44 ›› Issue (4): 667-676.DOI: 10.11996/JG.j.2095-302X.2023040667

• 图像处理与计算机视觉 • 上一篇下一篇

基于TR-YOLOv5的输电线路多类缺陷目标检测方法

郝帅(), 赵新生, 马旭(), 张旭, 何田, 侯李祥

西安科技大学电气与控制工程学院，陕西西安 710054

收稿日期:2023-01-31 接受日期:2023-03-16 出版日期:2023-08-31 发布日期:2023-08-16
通讯作者: 马旭(1985-)，女，讲师，博士。主要研究方向为图像处理和目标检测等。E-mail：414548542@qq.com
作者简介:
郝帅(1986-)，男，副教授，博士。主要研究方向为电气设备故障诊断和目标检测等。E-mail：haoxust@163.com
基金资助:
国家自然科学基金项目(51804250);中国博士后科学基金项目(2019M653874);中国博士后科学基金项目(2020M683522);陕西省科技计划项目(2021JQ-572);陕西省科技计划项目(2020JQ-757);陕西省教育厅科研计划项目(18JK0512);陕西省教育厅科研计划项目(17JK0503);西安市碑林区科技计划项目(GX2116)

Multi-class defect target detection method for transmission lines based on TR-YOLOv5

HAO Shuai(), ZHAO Xin-sheng, MA Xu(), ZHANG Xu, HE Tian, HOU Li-xiang

School of Electrical and Control Engineering, Xi’an University of Science and Technology, Xi’an Shaanxi 710054, China

Received:2023-01-31 Accepted:2023-03-16 Online:2023-08-31 Published:2023-08-16
Contact: MA Xu (1985-), lecturer, Ph.D. Her main research interests cover image processing, object detection, etc. E-mail：414548542@qq.com
About author:
HAO Shuai (1986-), associate professor, Ph.D. His main research interests cover electrical equipment fault diagnosis, target detection, etc. E-mail：haoxust@163.com
Supported by:
National Natural Science Foundation of China(51804250);China Postdoctoral Science Foundation(2019M653874);China Postdoctoral Science Foundation(2020M683522);Shaanxi Provincial Science and Technology Plan Project(2021JQ-572);Shaanxi Provincial Science and Technology Plan Project(2020JQ-757);Shaanxi Provincial Education Department Scientific Research Program(18JK0512);Shaanxi Provincial Education Department Scientific Research Program(17JK0503);Xi’an Beilin District Science and Technology Plan Project(GX2116)

摘要/Abstract

摘要：

针对复杂环境中输电线路多类缺陷目标的多尺度检测问题，提出一种基于Transformer和感受野模块的YOLOv5输电线路多类缺陷目标检测算法，简记为TR-YOLOv5。首先，搭建了YOLOv5网络，针对复杂背景造成缺陷目标的显著性低，进而影响检测精度的问题，在Backbone部分引入Transformer模块，通过利用多头注意力结构获取特征图像素点间的相关性和全局信息，增强缺陷目标的特征表达能力，从而提升模型检测精度；其次，由于待检测目标受多尺度影响，在Neck部分引入感受野模块提取目标不同尺度的特征，利用空洞卷积增大感受野，为后续PANet结构保留更细致的特征，增强Neck特征融合能力，提高模型对多尺度缺陷目标的检测精度；然后，为了提升预测边框回归精度，引入CIOU函数，进一步提高算法检测精度；最后，利用某电力巡检部门近3年的数据对该算法进行验证。实验结果表明，相比于7种对比算法，本文算法具有较高检测精度的同时具有较好的实时性，其平均检测精度可达95.6%，1280×720分辨率的巡检图像检测速度为125帧/秒。

关键词: YOLOv5, 输电线路缺陷检测, 空洞卷积, Transformer, 感受野模块, 损失函数

Abstract:

To address the problem of multi-scale detection of multi-class defect targets in transmission lines in complex environments, a YOLOv5 transmission line multi-class defect target detection algorithm was proposed based on Transformer and perceptual field modules, abbreviated as TR-YOLOv5. First, a YOLOv5 network was built to address the problem of low saliency of defect targets caused by complex backgrounds, which hindered accurate detection. The Transformer module was introduced in the Backbone part. By utilizing a multi-head attention structure to capture the correlations and global information between the pixels of feature maps, the feature expression capability of the defect targets was enhanced, thereby improving the detection accuracy of the model. Secondly, since the target being detected is impacted by multiple scales, a perceptual field module was introduced in the Neck part to extract features of different scales of the target. Null convolution was also employed to increase the perceptual field, while more detailed features were reserved for the subsequent PANet structure. Furthermore, the Neck feature fusion capability was bolstered to enhance the detection accuracy of the model for multi-scale defective targets. In addition, to enhance the precision of predicted border regression, the CIOU function was introduced to further boost the detection accuracy of the algorithm. Finally, the proposed algorithm was validated using the data of a power inspection department for the past three years. The experimental results demonstrated that the proposed algorithm could surpass seven comparative algorithms in terms of detection accuracy and real-time performance, with an average detection accuracy of 95.6% and the inspection image detection speed for 1280×720 resolution reaching 125 frames/second.

Key words: YOLOv5, transmission line defect detection, dilation convolution, Transformer, receptive field block, loss function

中图分类号:

TP391

郝帅, 赵新生, 马旭, 张旭, 何田, 侯李祥. 基于TR-YOLOv5的输电线路多类缺陷目标检测方法[J]. 图学学报, 2023, 44(4): 667-676.

HAO Shuai, ZHAO Xin-sheng, MA Xu, ZHANG Xu, HE Tian, HOU Li-xiang. Multi-class defect target detection method for transmission lines based on TR-YOLOv5[J]. Journal of Graphics, 2023, 44(4): 667-676.

图/表 13

图1 YOLOv5检测框架

Fig. 1 The detection framework of YOLOv5

图2 TR-YOLOv5检测算法框图

Fig. 2 Block diagram of the TR-YOLOv5 detection algorithm

图3 C3Trans模块和Transformer模块结构

Fig. 3 C3Trans module and Transformer module structure

图4 注意力机制计算流程

Fig. 4 Attention mechanism calculation process

图5 RFB网络结构

Fig. 5 RFB network structure

图6 缺陷目标((a)销钉脱落；(b)螺母松动；(c)螺母脱落；(d)放电间隙过大)

Fig. 6 Defect targets ((a) Pin off; (b) Loose nuts; (c) Nuts off; (d) Excessive discharge gap)

表1 缺陷名称和数量

Table 1 Defect name and quantity

缺陷类型	自定义名称	数量(张)
销钉脱落	XT	342
螺母松动	LSS	316
螺母脱落	LST	347
放电间隙过大	FDG	310
绝缘子脱落	JYT	212
鸟巢	NC	235
放电间隙设计问题	FDSJ	262
防鸟刺损坏	NS	231
放电间隙短接	FDG	310
垫片缺失	DP	323
间隔棒断开	JGB	364
杂物	ZW	248

图7 损失函数曲线

Fig. 7 Loss function curve

图8 mAP@0.5对比曲线

Fig. 8 mAP@0.5 contrast curve

表2 TR-YOLOv5模型各类目标检测结果(%)

Table 2 Detection result of TR-YOLOv5 model on various targets (%)

缺陷类型	Precision	Recall
XT	96.2	99.6
LSS	78.2	95.7
LST	97.5	98.9
FDG	96.1	97.3
JYT	99.3	97.1
NC	99.8	99.0
FDSJ	99.6	99.5
NS	96.3	96.8
FDD	98.0	98.6
DP	88.4	86.1
JGB	99.1	99.4
ZW	98.5	97.4
ALL	95.6	97.3

表3 消融实验

Table 3 Ablation experiment

C3Trans	RFB	CIOU	mAP (%)	FPS	参数量
-	-	-	95.0	131.0	7 042 489
√	-	-	95.3(↑0.3)	130.8	7 043 257
√	√	-	95.5(↑0.2)	125.0	7 704 057
√	√	√	95.6(↑0.1)	125.0	7 704 057

图9 不同算法检测结果((a)原始图像；(b) Faster RCNN；(c) SSD；(d) YOLOv3；(e) YOLOv4；(f) YOLOv5；(g) YOLOX-s；(h) YOLOv7；(i) TR-YOLOv5)

Fig. 9 The detection results of different algorithm ((a) Original images; (b) Faster RCNN; (c) SSD; (d) YOLOv3; (e) YOLOv4; (f) YOLOv5; (g) YOLOX-s; (h) YOLOv7; (i) TR-YOLOv5)

表4 不同检测算法性能对比

Table 4 Performance comparison of different algorithms

检测算法	FPS	mAP@0.5(%)	Weights(M)
Faster RCNN	2	94.9	113.8
SSD	29	91.7	96.0
YOLOv3	7	93.7	123.5
YOLOv4	14	94.3	256.2
YOLOv5	131	95.0	15.2
YOLOX-s	42	94.5	34.4
YOLOv7	36	95.3	74.8
TR-YOLOv5	125	95.6	15.8

参考文献 21

[1]	赵振兵, 齐鸿雨, 聂礼强. 基于深度学习的输电线路视觉检测研究综述[J]. 广东电力, 2019, 32(9): 11-23.
	ZHAO Z B, QI H Y, NIE L Q. Research overview on visual detection of transmission lines based on deep learning[J]. Guangdong Electric Power, 2019, 32(9): 11-23 (in Chinese).
[2]	蒲天骄, 乔骥, 韩笑, 等. 人工智能技术在电力设备运维检修中的研究及应用[J]. 高电压技术, 2020, 46(2): 369-383.
	PU T J, QIAO J, HAN X, et al. Research and application of artificial intelligence in operation and maintenance for power equipment[J]. High Voltage Engineering, 2020, 46(2): 369-383 (in Chinese).
[3]	WONG S Y, CHOE C W C, GOH H H, et al. Power transmission line fault detection and diagnosis based on artificial intelligence approach and its development in UAV: a review[J]. Arabian Journal for Science and Engineering, 2021, 46(10): 9305-9331. DOI
[4]	郝帅, 马瑞泽, 赵新生, 等. 基于卷积块注意模型的YOLOv3输电线路故障检测方法[J]. 电网技术, 2021, 45(8): 2979-2987.
	HAO S, MA R Z, ZHAO X S, et al. Fault detection of YOLOv3 transmission line based on convolutional block attention model[J]. Power System Technology, 2021, 45(8): 2979-2987 (in Chinese).
[5]	高文婷, 刘越. 面向移动增强现实的实时深度学习目标检测方法综述[J]. 图学学报, 2021, 42(4): 525-534.
	GAO W T, LIU Y. Review of real-time deep learning-based object detection for mobile augmented reality[J]. Journal of Graphics, 2021, 42(4): 525-534 (in Chinese).
[6]	李洪安, 郑峭雪, 陶若霖, 等. 基于深度学习的图像超分辨率研究综述[J]. 图学学报, 2023, 44(1): 1-15.
	LI H A, ZHENG Q X, TAO R L, et al. Review of image super-resolution based on deep learning[J]. Journal of Graphics, 2023, 44(1): 1-15 (in Chinese).
[7]	LIANG H G, ZUO C, WEI W M. Detection and evaluation method of transmission line defects based on deep learning[J]. IEEE Access, 2020, 8: 38448-38458. DOI URL
[8]	李辉, 董燕, 刘祥, 等. 基于两阶段深度网络的输电线路异常目标检测方法[J]. 控制与决策, 2022, 37(7): 1873-1882.
	LI H, DONG Y, LIU X, et al. Transmission line abnormal object detection method based on deep network of two-stage[J]. Control and Decision, 2022, 37(7): 1873-1882 (in Chinese).
[9]	REN S Q, HE K M, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137-1149. DOI PMID
[10]	戚银城, 武学良, 赵振兵, 等. 嵌入双注意力机制的Faster R-CNN航拍输电线路螺栓缺陷检测[J]. 中国图象图形学报, 2021, 26(11): 2594-2604.
	QI Y C, WU X L, ZHAO Z B, et al. Bolt defect detection for aerial transmission lines using Faster R-CNN with an embedded dual attention mechanism[J]. Journal of Image and Graphics, 2021, 26(11): 2594-2604 (in Chinese).
[11]	顾超越, 李喆, 史晋涛, 等. 基于改进Faster-RCNN的无人机巡检架空线路销钉缺陷检测[J]. 高电压技术, 2020, 46(9): 3089-3096.
	GU C Y, LI Z, SHI J T, et al. Detection for pin defects of overhead lines by UAV patrol image based on improved Faster-RCNN[J]. High Voltage Engineering, 2020, 46(9): 3089-3096 (in Chinese).
[12]	LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot multibox detector[C]// European Conference on Computer Vision. Cham: Springer International Publishing, 2016: 21-37.
[13]	REDMON J, FARHADI A. YOLOv3:an incremental improvement[EB/OL]. (2018-05-25) [2022-06-14]. https://arxiv.org/abs/1804.02767.
[14]	BOCHKOVSKIY A, WANG C Y, LIAO H Y. YOLOv4: optimal speed and accuracy of object detection[EB/OL]. (2020-06-23) [2022-01-04]. https://arxiv.org/abs/2004.10934.
[15]	翁智, 程曦, 郑志强. 基于改进YOLOv3的高压输电线路关键部件检测方法[J]. 计算机应用, 2020, 40(S2): 183-187.
	WENG Z, CHENG X, ZHENG Z Q. Detection method of key components of high-voltage transmission lines based on improved YOLOv3[J]. Journal of Computer Applications, 2020, 40(S2): 183-187 (in Chinese).
[16]	郝帅, 杨磊, 马旭, 等. 基于注意力机制与跨尺度特征融合的YOLOv5输电线路故障检测[J]. 中国电机工程学报, 2023, 43(6): 2319-2331.
	HAO S, YANG L, MA X, et al. Fault detection of YOLOv5 transmission lines based on attention mechanism and cross-scale feature fusion[J]. Proceedings of the CSEE, 2023, 43(6): 2319-2331 (in Chinese).
[17]	祝星馗, 蒋球伟. 基于CNN与Transformer的无人机图像目标检测研究[J]. 武汉理工大学学报: 信息与管理工程版, 2022, 44(2): 323-331.
	ZHU X K, JIANG Q W. Research on object detection for UAV images based on CNN and Transformer[J]. Journal of Wuhan University of Technology: Information & Management Engineering, 2022, 44(2): 323-331 (in Chinese).
[18]	郑伟, 赵金芳, 张奕婧, 等. 基于感受野扩增和注意力机制的U-Net脑肿瘤MR图像分割[J]. 河北大学学报: 自然科学版, 2022, 42(5): 542-551.
	ZHENG W, ZHAO J F, ZHANG Y J, et al. Segmentation algorithm of brain tumor MR images based on receptor field amplification and attentional mechanism[J]. Journal of Hebei University: Natural Science Edition, 2022, 42(5): 542-551 (in Chinese).
[19]	DU S J, ZHANG B F, ZHANG P, et al. An improved bounding box regression loss function based on CIOU loss for multi-scale object detection[C]// 2021 IEEE 2nd International Conference on Pattern Recognition and Machine Learning. New York: IEEE Press, 2021: 92-98.
[20]	GE Z, LIU S, WANG F, et al. YOLOX: exceeding YOLO series in 2021[EB/OL]. (2021-08-06) [2022-06-12]. https://doi.org/10.48550/arXiv.2107.08430.
[21]	WANG C, BOCHKOVSKIY A, LIAO H Y M. YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[EB/OL]. (2022-07-06) [2022-06-12]. https://doi.org/10.48550/arXiv.2207.02696.

基于TR-YOLOv5的输电线路多类缺陷目标检测方法

Multi-class defect target detection method for transmission lines based on TR-YOLOv5

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 13

参考文献 21

相关文章 15

编辑推荐

Metrics

本文评价

[1]	杨陈成 , 董秀成 , 侯兵 , 张党成 , 向贤明 , 冯琪茗 . 基于参考的Transformer纹理迁移深度图像超分辨率重建 [J]. 图学学报, 2023, 44(5): 861-867.
[2]	皮骏, 牛厚兴, 高志云. 融合 CA-BiFPN 的轻量化人体姿态估计算法 [J]. 图学学报, 2023, 44(5): 868-878.
[3]	杨红菊, 高敏, 张常有, 薄文, 武文佳, 曹付元. 一种面向图像修复的局部优化生成模型 [J]. 图学学报, 2023, 44(5): 955-965.
[4]	李利霞, 王鑫, 王军, 张又元. 基于特征融合与注意力机制的无人机图像小目标检测算法[J]. 图学学报, 2023, 44(4): 658-666.
[5]	胡欣, 周运强, 肖剑, 杨杰. 基于改进YOLOv5的螺纹钢表面缺陷检测[J]. 图学学报, 2023, 44(3): 427-437.
[6]	李刚, 张运涛, 汪文凯, 张东阳. 采用DETR与先验知识融合的输电线路螺栓缺陷检测方法[J]. 图学学报, 2023, 44(3): 438-447.
[7]	毛爱坤, 刘昕明, 陈文壮, 宋绍楼. 改进YOLOv5算法的变电站仪表目标检测方法[J]. 图学学报, 2023, 44(3): 448-455.
[8]	孙龙飞, 刘慧, 杨奉常, 李攀. 面向医学图像层间插值的循环生成网络研究[J]. 图学学报, 2023, 44(3): 502-512.
[9]	熊举举, 徐杨, 范润泽, 孙少聪. 基于轻量化视觉Transformer的花卉识别[J]. 图学学报, 2023, 44(2): 271-279.
[10]	陈刚, 张培基, 龚冬冬, 于俊清. 火电厂监控视频安全服检测方法研究[J]. 图学学报, 2023, 44(2): 291-297.
[11]	曹义亲, 伍铭林, 徐露. 基于改进YOLOv5算法的钢材表面缺陷检测[J]. 图学学报, 2023, 44(2): 335-345.
[12]	张伟康, 孙浩, 陈鑫凯, 李叙兵, 姚立纲, 东辉. 基于改进YOLOv5的智能除草机器人蔬菜苗田杂草检测研究[J]. 图学学报, 2023, 44(2): 346-356.
[13]	李小波, 李阳贵, 郭宁, 范震. 融合注意力机制的YOLOv5口罩检测算法[J]. 图学学报, 2023, 44(1): 16-25.
[14]	皮骏, 刘宇恒, 李久昊. 基于YOLOv5s的轻量化森林火灾检测算法研究[J]. 图学学报, 2023, 44(1): 26-32.
[15]	张倩, 王夏黎, 王炜昊, 武历展, 李超. 基于多尺度特征融合的细胞计数方法[J]. 图学学报, 2023, 44(1): 41-49.