基于PCSA-YOLOv7 Former的输电线路连接金具及其锈蚀检测方法

doi:10.11930/j.issn.1004-9649.202305035

中国电力 ›› 2024, Vol. 57 ›› Issue (6): 141-152.DOI: 10.11930/j.issn.1004-9649.202305035

基于PCSA-YOLOv7 Former的输电线路连接金具及其锈蚀检测方法

宋智伟¹(), 黄新波¹^,²(), 纪超¹, 张凡¹, 张烨¹

1. 西安工程大学电子信息学院，陕西西安　710048
2. 陕西省工业和信息化厅，陕西西安　710006

收稿日期:2023-05-08 接受日期:2024-01-15 出版日期:2024-06-28 发布日期:2024-06-25
作者简介:宋智伟（1998—），男，硕士，从事基于图像处理和深度学习的故障检测研究，E-mail：2796575906@qq.com
黄新波（1975—），男，通信作者，博士，二级教授，博士生导师，从事在线监测技术、图像识别技术和无线网络传感器研究，E-mail：huangxb1975@163.com
基金资助:
西安市科技计划项目（2022JH-RYFW-0031，22GXFW0041）；陕西省科学技术协会青年人才托举计划项目（20220133）；金属成形技术与重型装备全国重点实验室项目（S2208100.W03）。

Transmission Line Connection Fittings and Corrosion Detection Method Based on PCSA-YOLOv7 Former

Zhiwei SONG¹(), Xinbo HUANG¹^,²(), Chao JI¹, Fan ZHANG¹, Ye ZHANG¹

1. School of Electronics and Information, Xi'an Polytechnic University, Xi'an 710048, China
2. Department of Industry and Information Technology of Shaanxi Province, Xi'an 710006, China

Received:2023-05-08 Accepted:2024-01-15 Online:2024-06-28 Published:2024-06-25
Supported by:
This work is supported by Science & Technology Plan Project in Xi'an City (No.2022JH-RYFW-0031, No.22GXFW0041), Young Talent Fund of Association for Science & Technology in Shaanxi Province (No.20220133) and Program of National Key Laboratory of Metal Forming Technology and Heavy Equipment (No.S2208100.W03).

摘要/Abstract

摘要：

输电线路分布情况复杂且故障难以有效检测，其中连接金具长期暴露于复杂环境下易受到恶劣环境的影响出现锈蚀等故障。针对输电线路连接金具部件具有尺度多样性和存在着锈蚀故障检测精度低的问题，提出了一种基于双重注意力嵌入重构和Swin Transformer的输电线路连接金具组件及其锈蚀故障检测方法：PCSA-YOLOv7 Former。实验结果表明：该方法在构建的TLCF数据集上的综合检测性能领先于12类当前先进的目标检测算法，其中在测试集上的mAP_0.5达到94.9%，该方法相比于基线模型YOLOv7，其F1和mAP_0.5指标分别提升了2.6个百分点和2.2个百分点，说明该方法能够更全面地理解输电线路连接金具图像中的多尺度语义信息并学习到不易区分的微小细节表征。

关键词: 输电线路连接金具, PCSA-YOLOv7 Former, 双重注意力嵌入, Swin Transformer, 空洞空间金字塔池化

Abstract:

The transmission lines are complex in distribution and it is difficult to effectively detect their faults. Among them, the connecting fittings are susceptible to corrosion and other faults due to their long exposure to complex environments. Aiming at the problem that the transmission line connection fitting components are varied in scale and have poor accuracy in detecting their corrosion faults, a detection method is proposed for transmission line connection fittings and their corrosion faults based on dual attention embedding reconstruction and Swin Transformer, i.e., PCSA-YOLOv7 Former. The experimental results show that the proposed method is superior to 12 existing state-of-the-art object detection algorithms in comprehensive detection performance of the constructed TLCF dataset, with the mAP_0.5 of the test set reaching 94.9 %. Compared with the baseline model YOLOv7, the proposed method improves the indexes F1 and mAP_0.5 by 2.6 percentage points and 2.2 percentage points, respectively, indicating that the proposed method can more comprehensively understand the multi-scale semantic information in the images of transmission line connection fittings and learn their subtle details that are difficult to distinguish.

Key words: transmission line connection fittings, PCSA-YOLOv7 Former, dual attention embedding, Swin Transformer, atrous spatial pyramid pooling

宋智伟, 黄新波, 纪超, 张凡, 张烨. 基于PCSA-YOLOv7 Former的输电线路连接金具及其锈蚀检测方法[J]. 中国电力, 2024, 57(6): 141-152.

Zhiwei SONG, Xinbo HUANG, Chao JI, Fan ZHANG, Ye ZHANG. Transmission Line Connection Fittings and Corrosion Detection Method Based on PCSA-YOLOv7 Former[J]. Electric Power, 2024, 57(6): 141-152.

导出引用管理器 EndNote|Ris|BibTeX

链接本文: https://www.electricpower.com.cn/CN/10.11930/j.issn.1004-9649.202305035

https://www.electricpower.com.cn/CN/Y2024/V57/I6/141

图/表 15

图 1 PCSA-YOLOv7 Former模型结构示意

Fig.1 Model structure diagram of PCSA-YOLOv7 Former

图 2 ASPP 模型结构示意

Fig.2 Structure diagram of ASPP model

图 3 PCSA双重注意力嵌入模型

Fig.3 PCSA dual attention embedding model

图 4 Swin Transformer结构示意

Fig.4 Structure diagram of Swin Transformer

图 5 2阶段目标远近景示意

Fig.5 Far and near images of two-stage targets

图 6 主干特征提取可视化

Fig.6 Visualization of backbone feature extraction

图 7 验证结果分类混淆矩阵

Fig.7 Validation results classification confusion matrix

表 1 本文算法的检测结果

Table 1 Detection results of the proposed method

类别	P	R	F₁	mAP_0.5
U形挂环 (US)	0.962	0.867	0.912	0.938
三角联板 (TYP)	0.936	0.989	0.962	0.993
调整板 (ADP)	0.867	0.932	0.898	0.963
耐张线夹 (TEC)	0.785	0.836	0.811	0.828
紧固螺栓 (CLB)	0.855	0.967	0.908	0.968
绝缘子 (In)	0.998	1.000	0.998	0.996
所有 (All)	0.902	0.932	0.917	0.949

图 8 模型测试过程P-R曲线

Fig.8 P-R curves of model test

表 2 模型特征提取热力图可视化结果

Table 2 Heat map visualization results of model feature extraction

场景	原始图片	TYP	In	ADP	TEC	CLB	US
1
2
3
4

表 3 本文方法和其他先进算法的测试性能对比

Table 3 Test performance comparison between the proposed method and other advanced algorithms

算法	P/%	R/%	F₁/%	mAP_0.5/ %	GFLOPS/ G	Params/ M	FPS
Faster R- CNN	61.35	88.26	72.38	77.71	370.2	137.10	27.3
SSD	82.34	76.01	79.05	74.45	62.7	26.30	119.8
YOLOv3	91.62	60.77	73.07	82.76	66.2	61.90	63.5
YOLOv5	92.13	87.24	89.62	89.90	115.9	47.00	60.3
YOLOv5 Swin+ BiFPN	87.10	88.83	87.96	90.80	225.2	102.30	19.8
YOLOv5 Ghost backbone	90.20	92.10	91.14	93.90	42.2	24.20	44.8
FCOS	85.10	82.76	83.91	87.33	161.8	32.20	45.7
RetinaNet	91.10	76.00	82.87	77.85	170.1	37.90	44.3
CenterNet	98.80	63.26	77.13	87.18	70.2	32.70	88.4
TPH YOLOv5	88.11	89.60	88.85	92.50	270.3	112.90	16.2
YOLOX	89.20	92.29	90.72	92.40	26.6	8.94	60.1
YOLOv7	87.60	90.60	89.07	92.72	106.4	37.60	51.3
本文	90.23	93.20	91.69	94.90	114.3	41.40	47.1

图 9 不同算法测试过程mAP0.5曲线变化

Fig.9 mAP0.5 curves of different algorithms

表 4 不同算法检测结果比较

Table 4 Comparison of different algorithm results

算法	检测结果
算法	场景1	场景2	场景3
Faster R-CNN
YOLOv3
YOLOv5 Swin
FOS
CenterNet
YOLOv7
SSD
YOLOv5
YOLOv5 Ghost
RetinaNet
THP YOLOv5
本文

表 5 本文方法和基线方法在不同场景下锈蚀故障检测结果对比

Table 5 Comparison of rust detection results between the proposed method and the baseline method in different scenarios

锈蚀检测场景	原始图像	锈蚀故障检测结果		本文方法热力图识别结果
锈蚀检测场景	原始图像	本文方法	基线方法	本文方法热力图识别结果
1
2
3
4
5

表 6 消融实验对比结果

Table 6 Comparative results of ablation experiments

方法	P	R	F₁	mAP_0.5
YOLOv7（Baseline）	0.876	0.906	0.891	0.927
ASPP	0.896	0.902	0.909	0.921
PCSA	0.891	0.935	0.912	0.945
Swin	0.914	0.917	0.915	0.936
PCSA+ASPP	0.911	0.909	0.910	0.947
Swin+PCSA	0.908	0.912	0.912	0.941
本文	0.902	0.932	0.917	0.949

参考文献 25

1	刘传洋, 吴一全. 基于深度学习的输电线路视觉检测方法研究进展[J]. 中国电机工程学报, 2023, 43 (19): 7423- 7446.
	LIU Chuanyang, WU Yiquan. Research progress of vision detection methods based on deep learning for transmission lines[J]. Proceedings of the CSEE, 2023, 43 (19): 7423- 7446.
2	刘开培, 李博强, 秦亮, 等. 深度学习目标检测算法在架空输电线路绝缘子缺陷检测中的应用研究综述[J]. 高电压技术, 2023, 49 (9): 3584- 3595.
	LIU Kaipei, LI Boqiang, QIN Liang, et al. Review of application research of deep learning object detection algorithms in insulator defect detection of overhead transmission lines[J]. High Voltage Engineering, 2023, 49 (9): 3584- 3595.
3	姜岚, 程若恒, 唐波, 等. 基于图像边缘识别的输电线路拉线塔拉线张力测试方法[J]. 高电压技术, 2022, 48 (11): 4469- 4477.
	JIANG Lan, CHENG Ruoheng, TANG Bo, et al. Tension test method for guy cables of transmission line guyed tower based on image edge recognition[J]. High Voltage Engineering, 2022, 48 (11): 4469- 4477.
4	张烨, 高玉菡, 黄新波, 等. 基于Fine Mask RCNN的110~220kV输电铁塔涉鸟故障识别与评估[J]. 电网技术, 2022, 46 (6): 2132- 2144.
	ZHANG Ye, GAO Yuhan, HUANG Xinbo, et al. Fault identification and evaluation of tower wading in 110-220kV transmission line with fine mask RCNN[J]. Power System Technology, 2022, 46 (6): 2132- 2144.
5	RANI S, GHAI D, KUMAR S. Object detection and recognition using contour based edge detection and fast R-CNN[J]. Multimedia Tools and Applications, 2022, 81 (29): 42183- 42207. DOI
6	XU X Y, ZHAO M, SHI P X, et al. Crack detection and comparison study based on faster R-CNN and mask R-CNN[J]. Sensors, 2022, 22 (3): 1215. DOI
7	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 Yincheng, WU Xueliang, ZHAO Zhenbing, 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.
8	白洁音, 赵瑞, 谷丰强, 等. 多目标检测和故障识别图像处理方法[J]. 高电压技术, 2019, 45 (11): 3504- 3511.
	BAI Jieyin, ZHAO Rui, GU Fengqiang, et al. Multi-target detection and fault recognition image processing method[J]. High Voltage Engineering, 2019, 45 (11): 3504- 3511.
9	TRIPATHY S, SATPATHY M. SSD internal cache management policies: a survey[J]. Journal of Systems Architecture, 2022, 122, 102334. DOI
10	DEWI C, CHEN R C, JIANG X Y, et al. Deep convolutional neural network for enhancing traffic sign recognition developed on YOLO v4[J]. Multimedia Tools and Applications, 2022, 81 (26): 37821- 37845. DOI
11	WANG Q F, CHENG M, HUANG S, et al. A deep learning approach incorporating YOLO v5 and attention mechanisms for field real-time detection of the invasive weed Solanum rostratum dunal seedlings[J]. Computers and Electronics in Agriculture, 2022, 199, 107194. DOI
12	JIANG K L, XIE T Y, YAN R, et al. An attention mechanism-improved YOLOv7 object detection algorithm for hemp duck count estimation[J]. Agriculture, 2022, 12 (10): 1659. DOI
13	谢静, 杜耀文, 刘志坚, 等. 基于轻量化改进型YOLOv5s的可见光绝缘子缺陷检测算法[J]. 电网技术, 2023, 47 (12): 5273- 5283.
	XIE Jing, DU Yaowen, LIU Zhijian, et al. Defect detection algorithm based on lightweight and improved YOLOv5s for visible light insulators[J]. Power System Technology, 2023, 47 (12): 5273- 5283.
14	汝承印, 张仕海, 张子淼, 等. 基于轻量级MobileNet-SSD和MobileNetV2-DeeplabV³⁺的绝缘子故障识别方法[J]. 高电压技术, 2022, 48 (9): 3670- 3679.
	RU Chengyin, ZHANG Shihai, ZHANG Zimiao, et al. Fault identification method for high voltage power grid insulator based on lightweight MobileNet-SSD and MobileNetV2-DeeplabV³⁺ network[J]. High Voltage Engineering, 2022, 48 (9): 3670- 3679.
15	LIANG Y H, LIN Y, LU Q. Forecasting gold price using a novel hybrid model with ICEEMDAN and LSTM-CNN-CBAM[J]. Expert Systems with Applications, 2022, 206, 117847. DOI
16	HUANG J, REN L F, ZHOU X K, et al. An improved neural network based on SENet for sleep stage classification[J]. IEEE Journal of Biomedical and Health Informatics, 2022, 26 (10): 4948- 4956. DOI
17	HOU Q B, ZHOU D Q, FENG J S. Coordinate attention for efficient mobile network design[C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Nashville, TN, USA. IEEE, 2021: 13708–13717.
18	ZHAO Y L, JU Z J, SUN T A, et al. TGC-YOLOv5: an enhanced YOLOv5 drone detection model based on transformer, GAM & CA attention mechanism[J]. Drones, 2023, 7 (7): 446. DOI
19	VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems. Long Beach, California, USA. ACM, 2017: 6000–6010.
20	JIAO M, WANG D Q, QIU J L. A GRU-RNN based momentum optimized algorithm for SOC estimation[J]. Journal of Power Sources, 2020, 459, 228051. DOI
21	ZHA W S, LIU Y P, WAN Y J, et al. Forecasting monthly gas field production based on the CNN-LSTM model[J]. Energy, 2022, 260, 124889. DOI
22	SONG Z W, HUANG X B, JI C, et al. Double-attention YOLO: vision transformer model based on image processing technology in complex environment of transmission line connection fittings and rust detection[J]. Machines, 2022, 10 (11): 1002. DOI
23	GONG H, MU T K, LI Q X, et al. Swin-transformer-enabled YOLOv5 with attention mechanism for small object detection on satellite images[J]. Remote Sensing, 2022, 14 (12): 2861. DOI
24	宋智伟, 黄新波, 纪超, 等. 基于Flexible YOLOv7的输电线路绝缘子缺陷检测和故障预警方法[J/OL]. 高电压技术: 1–16[2023-08-01].https://doi.org/10.13336/j.1003-6520.hve.20221693.
	SONG Zhiwei, Huang Xinbo, Ji Chao et al. Transmission line insulator defect detection and fault warning method based on Flexible YOLOv7[J/OL]. High Voltage Technology : 1–16 [2023-08-01].https://doi.org/10.13336/j.1003-6520.hve.20221693.
25	LIU R R, TAO F, LIU X T, et al. RAANet: a residual ASPP with attention framework for semantic segmentation of high-resolution remote sensing images[J]. Remote Sensing, 2022, 14 (13): 3109. DOI

基于PCSA-YOLOv7 Former的输电线路连接金具及其锈蚀检测方法

Transmission Line Connection Fittings and Corrosion Detection Method Based on PCSA-YOLOv7 Former

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 15

参考文献 25

相关文章 0

编辑推荐

Metrics

模态框（Modal）标题

基于PCSA-YOLOv7 Former的输电线路连接金具及其锈蚀检测方法

Transmission Line Connection Fittings and Corrosion Detection Method Based on PCSA-YOLOv7 Former

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 15

参考文献 25

相关文章 0

编辑推荐

Metrics