基于改进YOLOv8的防振锤缺陷目标检测算法

doi:10.11996/JG.j.2095-302X.2025030532

图学学报 ›› 2025, Vol. 46 ›› Issue (3): 532-541.DOI: 10.11996/JG.j.2095-302X.2025030532

• 图像处理与计算机视觉 • 上一篇下一篇

基于改进YOLOv8的防振锤缺陷目标检测算法

牛杭(), 葛鑫雨, 赵晓瑜, 杨珂, 王乾铭(), 翟永杰

华北电力大学自动化系，河北保定 071003

收稿日期:2024-09-10 接受日期:2024-10-28 出版日期:2025-06-30 发布日期:2025-06-13
通讯作者:王乾铭(1995-)，男，讲师，博士。主要研究方向为电力视觉、深度学习以及故障诊断。E-mail：qianmingwang@ncepu.edu.cn
第一作者:牛杭(1991-)，男，讲师，博士。主要研究方向为智能检测、诊断与控制。E-mail：hangniu@ncepu.edu.cn
基金资助:
国家自然科学基金(62373151);中央高校基本科研业务费专项资金(2022MS097);河北省自然科学基金面上项目(F2023502010);河北省高等学校科学技术研究项目(QN2024071)

Vibration damper defect detection algorithm based on improved YOLOv8

NIU Hang(), GE Xinyu, ZHAO Xiaoyu, YANG Ke, WANG Qianming(), ZHAI Yongjie

Department of Automation, North China Electric Power University, Baoding Hebei 071003, China

Received:2024-09-10 Accepted:2024-10-28 Published:2025-06-30 Online:2025-06-13
Contact: WANG Qianming (1995-), lecturer, Ph.D. His main research interests cover power vision, deep learning and fault diagnosis. E-mail：qianmingwang@ncepu.edu.cn
First author：NIU Hang (1991-), lecturer, Ph.D. His main research interests cover intelligent detection, diagnosis and control. E-mail：hangniu@ncepu.edu.cn
Supported by:
The National Natural Science Foundation of China(62373151);The Fundamental Research Funds for the Central Universities(2022MS097);Natural Science Foundation of Hebei Province(F2023502010);Science and Technology Research Project of Hebei Colleges and Universities(QN2024071)

摘要/Abstract

摘要：

利用无人机对输电线路进行巡检的过程中，航拍的防振锤图像目标尺度多变且背景复杂，容易导致漏检、误检。为此，针对现有目标检测算法在复杂背景和多尺度目标检测方面的局限性，提出了一种基于改进YOLOv8的防振锤缺陷目标检测算法。首先，提出多尺度特征提取(MSFE)模块，有效扩大模型的感受野，增强模型的多尺度特征提取能力；其次，设计空间金字塔核注意力(SPKA)模块，提升模型对目标的全局感知能力，在多尺度特征融合过程中抑制复杂背景的干扰；最后，由于小尺度目标在图像中易被忽略，在网络中增加丰富小目标语义信息的特征层(STSIL)，提高小目标缺陷检测能力。对比实验中，改进算法与基线模型YOLOv8s相比mAP⁵⁰提升了5.7%，防振锤正常、倾斜、掉落的AP⁵⁰分别提升了3.4%，4.5%和9.2%，证明了改进算法对防振锤缺陷检测的有效性与先进性，且其应用将有助于保障电力系统的安全可靠运行。

关键词: 输电线路, 防振锤, 缺陷检测, 多尺度目标, 注意力机制

Abstract:

During drone inspections of transmission lines, the aerial images of vibration dampers exhibited varying target scales and complex backgrounds, which can easily lead to missed or false detections. To address the limitations of existing object detection algorithms in handling complex backgrounds and multi-scale target detection, an improved YOLOv8-based detection algorithm for identifying defects in vibration dampers was proposed. Firstly, to enhance the model’s ability to extract multi-scale features, a multi-scale feature extraction (MSFE) module was introduced, effectively expanding the model’s receptive field. Secondly, to suppress interference from complex backgrounds during the multi-scale feature fusion process, a space pyramid kernel attention (SPKA) module was designed to improve the model’s global awareness of the target. Lastly, to improve the detection capability for small target defects, a small target semantic information layer (STSIL) was added to the network, providing rich semantic information for small-scale targets that were easily overlooked in images. In the comparison experiments, the mAP⁵⁰ of the improved algorithm increased by 5.7% over the baseline model YOLOv8s, with AP⁵⁰ for normal, tilted, and fallen vibration dampers increasing by 3.4%, 4.5%, and 9.2%, respectively, demonstrating the effectiveness and superiority of the proposed algorithm in detecting defects in vibration dampers. The application of the proposed algorithm was expected to contribute to ensuring the safe and reliable operation of the power system.

Key words: transmission lines, vibration dampers, defect detection, multi-scale targets, feature fusion, attention mechanism

中图分类号:

牛杭, 葛鑫雨, 赵晓瑜, 杨珂, 王乾铭, 翟永杰. 基于改进YOLOv8的防振锤缺陷目标检测算法[J]. 图学学报, 2025, 46(3): 532-541.

NIU Hang, GE Xinyu, ZHAO Xiaoyu, YANG Ke, WANG Qianming, ZHAI Yongjie. Vibration damper defect detection algorithm based on improved YOLOv8[J]. Journal of Graphics, 2025, 46(3): 532-541.

图/表 13

参考文献 26

[1]	GHASEMI Y, JEONG H, CHOI S H, et al. Deep learning- based object detection in augmented reality: a systematic review[J]. Computers in Industry, 2022, 139: 103661.
[2]	ZHAI Y J, YANG X, WANG Q M, et al. Hybrid knowledge R-CNN for transmission line multifitting detection[J]. IEEE Transactions on Instrumentation and Measurement, 2021, 70: 5013312.
[3]	赵振兵, 蒋志钢, 李延旭, 等. 输电线路部件视觉缺陷检测综述[J]. 中国图象图形学报, 2021, 26(11): 2545-2560.
	ZHAO Z B, JIANG Z G, LI Y X, et al. Overview of visual defect detection of transmission line components[J]. Journal of Image and Graphics, 2021, 26(11): 2545-2560 (in Chinese).
[4]	翟永杰, 赵晓瑜, 王璐瑶, 等. IDD-YOLOv7: 一种用于输电线路绝缘子多缺陷的轻量化检测方法[J]. 图学学报, 2024, 45(1): 90-101. DOI
	ZHAI Y J, ZHAO X Y, WANG L Y, et al. IDD-YOLOv7: a lightweight method for multiple defect detection of insulators in transmission lines[J]. Journal of Graphics, 2024, 45(1): 90-101 (in Chinese). DOI
[5]	BAO W X, REN Y X, LIANG D, et al. Defect detection algorithm of anti-vibration hammer based on improved cascade R-CNN[C]// 2020 International Conference on Intelligent Computing and Human-Computer Interaction. New York: IEEE Press, 2020: 294-297.
[6]	郝帅, 赵新生, 马旭, 等. 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. DOI
	HAO S, ZHAO X S, MA X, et al. Multi-class defect target detection method for transmission lines based on TR-YOLOv5[J]. Journal of Graphics, 2023, 44(4): 667-676 (in Chinese). DOI
[7]	ZHOU F R, WEN G, QIAN G C, et al. Retracted: a high-efficiency deep-learning-based antivibration hammer defect detection model for energy-efficient transmission line inspection systems[J]. International Journal of Antennas and Propagation, 2022, 2022(1): 3867581.
[8]	LIU Z Y, WU G P, HE W S, et al. Key target and defect detection of high-voltage power transmission lines with deep learning[J]. International Journal of Electrical Power & Energy Systems, 2022, 142: 108277.
[9]	梁华刚, 赵慧霞, 刘丽华, 等. 结合跨层特征融合与级联检测器的防震锤缺陷检测[J]. 中国图象图形学报, 2023, 28(11): 3485-3496.
	LIANG H G, ZHAO H X, LIU L H, et al. Combining cross-layer feature fusion with cascade detectors for anti-vibration hammer defects detection[J]. Journal of Image and Graphics, 2023, 28(11): 3485-3496 (in Chinese).
[10]	ZHAI Y J, YANG K, ZHAO Z Y, et al. Geometric characteristic learning R-CNN for shockproof hammer defect detection[J]. Engineering Applications of Artificial Intelligence, 2022, 116: 105429.
[11]	LAU K W, PO L M, REHMAN Y A U. Large separable kernel attention: Rethinking the large kernel attention design in CNN[J]. Expert Systems with Applications, 2024, 236: 121352.
[12]	LIN T Y, DOLLÁR P, GIRSHICK R, et al. Feature pyramid networks for object detection[C]// 2017 IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2017: 936-944.
[13]	CHOLLET F. Xception: deep learning with depthwise separable convolutions[C]// 2017 IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2017: 1800-1807.
[14]	YU F, KOLTUN V. Multi-scale context aggregation by dilated convolutions[EB/OL]. [2024-07-10]https://arxiv.org/abs/1511.07122.
[15]	VOOTURI D T, VARMA G, KOTHAPALLI K. Dynamic block sparse reparameterization of convolutional neural networks[C]// 2019 IEEE/CVF International Conference on Computer Vision Workshop. New York: IEEE Press, 2019: 3046-3053.
[16]	HUANG Z C, WANG J L, FU X S, et al. DC-SPP-YOLO: dense connection and spatial pyramid pooling based YOLO for object detection[J]. Information Sciences, 2020, 522: 241-258.
[17]	WANG P Q, CHEN P F, YUAN Y, et al. Understanding convolution for semantic segmentation[C]// 2018 IEEE Winter Conference on Applications of Computer Vision. New York: IEEE Press, 2018: 1451-1460.
[18]	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.
[19]	张相胜, 杨骁. 基于改进YOLOv7-tiny的橡胶密封圈缺陷检测方法[J]. 图学学报, 2024, 45(3): 446-453. DOI
	ZHANG X S, YANG X. Defect detection method of rubber seal ring based on improved YOLOv7-tiny[J]. Journal of Graphics, 2024, 45(3): 446-453 (in Chinese). DOI
[20]	LIN T Y, MAIRE M, BELONGIE S, et al. Microsoft COCO: common objects in context[C]// The 13th European Conference on Computer Vision. Cham: Springer, 2014: 740-755.
[21]	ZHU X K, LYU S C, WANG X, et al. TPH-YOLOv5: improved YOLOv5 based on transformer prediction head for object detection on drone-captured scenarios[C]// 2021 IEEE/CVF International Conference on Computer Vision. New York: IEEE Press, 2021: 2778-2788.
[22]	LI L L, WANG Z F, ZHANG T T. GBH-YOLOv5: ghost convolution with BottleneckCSP and tiny target prediction head incorporating YOLOv5 for PV panel defect detection[J]. Electronics, 2023, 12(3): 561.
[23]	LI C Y, LI L L, JIANG H L, et al. YOLOv6:a single-stage object detection framework for industrial applications[EB/OL]. [2024-07-10]https://arxiv.org/abs/2209.02976.
[24]	WANG A, CHEN H, LIU L H, et al. YOLOv10:real-time end-to-end object detection[EB/OL]. [2024-07-10]https://arxiv.org/abs/2405.14458.
[25]	LI K, WAN G, CHENG G, et al. Object detection in optical remote sensing images: a survey and a new benchmark[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2020, 159: 296-307.
[26]	VIEIRA-E-SILVA A L B, DE CASTRO FELIX H, DE MENEZES CHAVES T, et al.STN PLAD: a dataset for multi-size power line assets detection in high-resolution UAV images[C]// The 34th SIBGRAPI Conference on Graphics, Patterns and Images. New York: IEEE Press, 2021: 215-222.

检测算法	AP⁵⁰			mAP⁵⁰	mAP^50:95
检测算法	正常	倾斜	掉落	mAP⁵⁰	mAP^50:95
TPH-YOLOv5	68.9	55.3	70.0	64.7	37.0
GBH-YOLOv5	69.0	56.7	66.4	64.0	35.5
YOLOv6s	65.1	56.5	71.3	64.3	35.9
YOlOv7-tiny	66.6	53.3	77.3	65.7	36.0
YOLOv8s	69.3	55.3	69.8	64.8	37.2
YOLOv10s	68.1	51.7	71.3	63.7	34.7
YOLO11s	67.5	52.8	67.7	62.6	34.8
Ours	72.7	59.8	79.0	70.5	39.1

检测算法	AP⁵⁰			mAP⁵⁰	mAP^50:95
检测算法	正常	倾斜	掉落	mAP⁵⁰	mAP^50:95
TPH-YOLOv5	68.9	55.3	70.0	64.7	37.0
GBH-YOLOv5	69.0	56.7	66.4	64.0	35.5
YOLOv6s	65.1	56.5	71.3	64.3	35.9
YOlOv7-tiny	66.6	53.3	77.3	65.7	36.0
YOLOv8s	69.3	55.3	69.8	64.8	37.2
YOLOv10s	68.1	51.7	71.3	63.7	34.7
YOLO11s	67.5	52.8	67.7	62.6	34.8
Ours	72.7	59.8	79.0	70.5	39.1

实验	MSFE	SPKA	STSIL	AP^50:95/%			mAP⁵⁰/%	mAP^50:95/%	Params/M	FPS/(帧/秒)
实验	MSFE	SPKA	STSIL	Small	Medium	Large	mAP⁵⁰/%	mAP^50:95/%	Params/M	FPS/(帧/秒)
1				34.5	48.6	30.0	63.1	35.8	11.2	185.2
2	√			35.5	50.9	43.5	64.7	35.2	10.0	212.8
3		√		35.5	51.2	35.0	65.5	37.7	12.3	153.8
4			√	36.7	46.2	25.0	65.8	37.9	11.4	172.4
5	√	√		35.1	52.8	42.5	66.4	37.6	11.5	178.8
6	√		√	37.4	49.6	35.7	66.1	37.1	10.9	196.0
7		√	√	35.2	51.2	35.0	66.7	37.2	11.7	163.9
8	√	√	√	37.6	51.6	49.0	69.7	38.9	10.8	179.6

实验	MSFE	SPKA	STSIL	AP^50:95/%			mAP⁵⁰/%	mAP^50:95/%	Params/M	FPS/(帧/秒)
实验	MSFE	SPKA	STSIL	Small	Medium	Large	mAP⁵⁰/%	mAP^50:95/%	Params/M	FPS/(帧/秒)
1				34.5	48.6	30.0	63.1	35.8	11.2	185.2
2	√			35.5	50.9	43.5	64.7	35.2	10.0	212.8
3		√		35.5	51.2	35.0	65.5	37.7	12.3	153.8
4			√	36.7	46.2	25.0	65.8	37.9	11.4	172.4
5	√	√		35.1	52.8	42.5	66.4	37.6	11.5	178.8
6	√		√	37.4	49.6	35.7	66.1	37.1	10.9	196.0
7		√	√	35.2	51.2	35.0	66.7	37.2	11.7	163.9
8	√	√	√	37.6	51.6	49.0	69.7	38.9	10.8	179.6

检测算法	AP⁵⁰				mAP⁵⁰
检测算法	立交桥	运动场	飞机	油桶	mAP⁵⁰
TPH-YOLOv5	93.1	99.2	94.4	97.5	96.1
GBH-YOLOv5	89.3	99.3	94.6	96.9	95.0
YOLOv6s	77.3	98.4	90.7	95.4	90.5
YOlOv7-tiny	94.0	98.8	94.3	96.9	96.0
YOLOv8s	76.2	99.2	94.4	97.2	91.7
YOLOv10s	87.7	98.2	94.4	98.0	94.6
YOLO11s	94.4	98.5	95.0	97.0	96.2
Ours	94.7	99.5	95.5	97.6	96.8

基于改进YOLOv8的防振锤缺陷目标检测算法

Vibration damper defect detection algorithm based on improved YOLOv8

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 13

参考文献 26

相关文章 15

编辑推荐

Metrics

本文评价

检测算法	AP⁵⁰					mAP⁵⁰
检测算法	杆塔	绝缘子	间隔棒	防振锤	塔牌	mAP⁵⁰
TPH-YOLOv5	41.9	81.3	90.4	25.8	81.9	64.2
GBH-YOLOv5	34.7	75.1	85.8	27.6	80.8	60.8
YOLOv6s	43.9	88.3	77.5	8.3	73.4	58.3
YOlOv7-tiny	37.6	72.6	79.8	13.8	71.8	55.1
YOLOv8s	46.4	78.5	91.8	27.1	78.2	64.4
YOLOv10s	39.2	89.0	91.4	24.1	73.9	63.5
YOLO11s	42.7	88.7	88.0	25.4	83.9	65.7
Ours	42.9	89.2	92.5	29.2	82.2	67.2

检测算法	mAP⁵⁰
检测算法	正常图像	有雨图像	有雾图像	暗化图像	模糊图像
TPH-YOLOv5	64.7	59.5(↓ 5.2)	56.2(↓ 8.5)	59.3(↓ 5.4)	57.7(↓ 7.0)
GBH-YOLOv5	64.0	59.2(↓ 4.8)	58.9(↓ 5.1)	57.1(↓ 6.9)	56.1(↓ 7.9)
YOLOv6s	64.3	58.7(↓ 5.6)	57.9(↓ 6.4)	59.8(↓ 4.5)	56.2(↓ 8.1)
YOlOv7-tiny	65.7	56.5(↓ 9.2)	61.9(↓ 3.8)	57.8(↓ 7.9)	58.8(↓ 6.9)
YOLOv8s	64.8	58.4(↓ 6.4)	56.9(↓ 7.9)	55.7(↓ 9.1)	57.2(↓ 7.6)
YOLOv10s	63.7	57.7(↓ 6.0)	58.4(↓ 5.3)	59.1(↓ 4.6)	58.7(↓ 5.0)
YOLO11s	62.6	56.7(↓ 5.9)	55.9(↓ 6.7)	57.3(↓ 5.3)	55.1(↓ 7.5)
Ours	70.5	66.7(↓ 3.8)	67.2(↓ 3.3)	67.9(↓ 2.6)	67.7(↓ 2.8)

[1]	王素琴, 杜雨洁, 石敏, 朱登明. 类别不均衡的少样本工业产品表观缺陷检测[J]. 图学学报, 2025, 46(3): 568-577.
[2]	于冰, 程广, 黄东晋, 丁友东. 基于双流网络融合的三维人体网格重建[J]. 图学学报, 2025, 46(3): 625-634.
[3]	雷玉林, 刘利刚. 基于深度强化学习的可缓冲的物体运输和装箱[J]. 图学学报, 2025, 46(3): 697-708.
[4]	王志东, 陈晨阳, 刘晓明. 基于自适应特征提取的通信光缆缺陷检测方法[J]. 图学学报, 2025, 46(2): 241-248.
[5]	张立立, 杨康, 张珂, 魏薇, 李晶, 谭洪鑫, 张翔宇. 面向柴油车辆排放黑烟的改进型YOLOv8检测算法研究[J]. 图学学报, 2025, 46(2): 249-258.
[6]	翟永杰, 王璐瑶, 赵晓瑜, 胡哲东, 王乾铭, 王亚茹. 基于级联查询-位置关系的输电线路多金具检测方法[J]. 图学学报, 2025, 46(2): 288-299.
[7]	赵振兵, 韩钰, 唐辰康. 基于改进YOLOv8的配电线路绝缘子缺陷级联检测方法[J]. 图学学报, 2025, 46(1): 1-12.
[8]	崔克彬, 耿佳昌. 基于EE-YOLOv8s的多场景火灾迹象检测算法[J]. 图学学报, 2025, 46(1): 13-27.
[9]	王志东, 陈晨阳, 刘晓明. 基于轻量化改进YOLOv8的通信光缆缺陷检测[J]. 图学学报, 2025, 46(1): 28-34.
[10]	陈冠豪, 徐丹, 贺康建, 施洪贞, 张浩. 基于转置注意力和CNN的图像超分辨率重建网络[J]. 图学学报, 2025, 46(1): 35-46.
[11]	张文祥, 王夏黎, 王欣仪, 杨宗宝. 一种强化伪造区域关注的深度伪造人脸检测方法[J]. 图学学报, 2025, 46(1): 47-58.
[12]	苑朝, 赵明雪, 张丰羿, 冯晓勇, 李冰, 陈瑞. 基于点云特征增强的复杂室内场景3D目标检测[J]. 图学学报, 2025, 46(1): 59-69.
[13]	卢洋, 陈林慧, 姜晓恒, 徐明亮. SDENet：基于多尺度注意力质量感知的合成缺陷数据评价网络[J]. 图学学报, 2025, 46(1): 94-103.
[14]	李琼, 考月英, 张莹, 徐沛. 面向无人机航拍图像的目标检测研究综述[J]. 图学学报, 2024, 45(6): 1145-1164.
[15]	胡凤阔, 叶兰, 谭显峰, 张钦展, 胡志新, 方清, 王磊, 满孝锋. 一种基于改进YOLOv8的轻量化路面病害检测算法[J]. 图学学报, 2024, 45(5): 892-900.

检测算法	AP⁵⁰			mAP⁵⁰	mAP^50:95
检测算法	正常	倾斜	掉落	mAP⁵⁰	mAP^50:95
YOLOv8n	62.8	54.1	66.5	61.1	33.7
YOLOv8m	70.8	56.1	72.1	66.3	37.4
YOLOv8l	71.0	57.1	75.2	67.8	37.9
YOLOv8n-ours	68.9(↑6.1)	55.2(↑1.1)	70.0(↑3.5)	64.7(↑3.6)	37.0(↑3.3)
YOLOv8m-ours	70.9(↑0.1)	57.6(↑1.5)	74.1(↑2.0)	67.5(↑1.2)	38.4(↑1.0)
YOLOv8l-ours	73.2(↑2.2)	57.9(↑0.8)	76.6(↑1.4)	69.2(↑1.4)	42.2(↑4.3)

检测算法	AP⁵⁰			mAP⁵⁰	mAP^50:95
检测算法	正常	倾斜	掉落	mAP⁵⁰	mAP^50:95
YOLOv8n	62.8	54.1	66.5	61.1	33.7
YOLOv8m	70.8	56.1	72.1	66.3	37.4
YOLOv8l	71.0	57.1	75.2	67.8	37.9
YOLOv8n-ours	68.9(↑6.1)	55.2(↑1.1)	70.0(↑3.5)	64.7(↑3.6)	37.0(↑3.3)
YOLOv8m-ours	70.9(↑0.1)	57.6(↑1.5)	74.1(↑2.0)	67.5(↑1.2)	38.4(↑1.0)
YOLOv8l-ours	73.2(↑2.2)	57.9(↑0.8)	76.6(↑1.4)	69.2(↑1.4)	42.2(↑4.3)