TFD-YOLOv8: a transmission line foreign object detection method

doi:10.11996/JG.j.2095-302X.2024050901

Abstract

Abstract:

Foreign object detection based on UAV aerial images is an important aspect for intelligent inspection of transmission lines. The YOLO target detection algorithm is high in accuracy and speed, making it the current mainstream algorithm. However, when carrying out transmission line foreign object detection, due to the variable scale and insignificant features of foreign object targets, problems such as misdetection and omission detection would arise. A YOLOv8 model (transmission line foreign detection-YOLOv8, TFD-YOLOv8) was proposed for transmission line foreign object detection. A two-branch downsampling module was constructed in the YOLOv8s neck network to intercept the scale-related detail information easily lost during the downsampling process, achieving the efficient fusion of semantic and detail information and improving the information consistency of feature maps at different scales. Then, a mix-enhancement attention module was inserted into the backbone network to simultaneously extract global and local features of the image, generating spatial attention and channel attention, respectively, and resulting in a mix-enhancement attention including local, global, spatial, and channel information. This enhanced the network’s ability to capture the key features of the targe. The experimental results showed that compared with the baseline model, the proposed method improved the average detection accuracy by 6.7%, and the accuracy and recall by 12.9% and 5.1%, respectively. This method demonstrated advantages in terms of detection accuracy and complexity compared with several existing target detection methods.

Key words: YOLOv8, transmission line, object detection, downsampling, mix-enhancement

CLC Number:

WANG Yaru, FENG Lilong, SONG Xiaoke, QU Zhuo, YANG Ke, WANG Qianming, ZHAI Yongjie. TFD-YOLOv8: a transmission line foreign object detection method[J]. Journal of Graphics, 2024, 45(5): 901-912.

Figures/Tables 13

References 31

[1]	翟永杰, 郭聪彬, 王乾铭, 等. 基于隐含空间知识融合的输电线路多金具检测方法[J]. 图学学报, 2023, 44(5): 918-927.
	ZHAI Y J, GUO C B, WANG Q M, et al. Multi-fitting detection method for transmission lines based on implicit spatial knowledge fusion[J]. Journal of Graphics, 2023, 44(5): 918-927 (in Chinese).
[2]	阎光伟, 刘润泽, 焦润海, 等. 基于改进Cascade RCNN的输电线路防振锤脱落检测方法[J]. 图学学报, 2023, 44(5): 849-860. DOI
	YAN G W, LIU R Z, JIAO R H, et al. Detection method of dropped anti-vibration hammer for transmission line based on improved Cascade RCNN[J]. Journal of Graphics, 2023, 44(5): 849-860 (in Chinese).
[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]	赵振兵, 熊静, 李冰, 等. 基于改进Cascade R-CNN的典型金具及其部分缺陷检测方法[J]. 高电压技术, 2022, 48(3): 1060-1067.
	ZHAO Z B, XIONG J, LI B, et al. Typical fittings and its partial defect detection method based on improved cascade R-CNN[J]. High Voltage Engineering, 2022, 48(3): 1060-1067 (in Chinese).
[5]	WU X, YUAN P, PENG Q, et al. Detection of bird nests in overhead catenary system images for high-speed rail[J]. Pattern Recognition, 2016, 51: 242-254.
[6]	LU J F, XU X Y, LI X, et al. Detection of bird’s nest in high power lines in the vicinity of remote campus based on combination features and cascade classifier[J]. IEEE Access, 2018, 6: 39063-39071.
[7]	YAO N, ZHU L F. A novel foreign object detection algorithm based on GMM and K-means for power transmission line inspection[J]. Journal of Physics: Conference Series, 2020, 1607(1): 012014.
[8]	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
[9]	CAI Z W, VASCONCELOS N. Cascade R-CNN: delving into high quality object detection[C]// 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2018: 6154-6162.
[10]	ZHANG H K, CHANG H, MA B P, et al. Dynamic R-CNN: towards high quality object detection via dynamic training[M]// Computer Vision - ECCV 2020. Cham: Springer International Publishing, 2020: 260-275.
[11]	BOCHKOVSKIY A, WANG C Y, LIAO H Y M. YOLOv4: optimal speed and accuracy of object detection[EB/OL]. [2024-01-12]. http://arxiv.org/abs/2004.10934.
[12]	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 Workshops. New York: IEEE Press, 2021: 2778-2788.
[13]	CHEN Q, WANG Y M, YANG T, et al. You only look one-level feature[C]// 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2021: 13034-13043.
[14]	LI C Y, LI L L, JIANG H L, et al. YOLOv6: a single-stage object detection framework for industrial applications[EB/OL]. [2023-12-03]. http://arxiv.org/abs/2209.02976.
[15]	WANG C Y, BOCHKOVSKIY A, LIAO H Y M. YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]// 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2023: 7464-7475.
[16]	REIS D, KUPEC J, HONG J, et al. Real-time flying object detection with YOLOv8[EB/OL]. [2023-12-13]. http://arxiv.org/abs/2305.09972.
[17]	QIU Z B, ZHU X, LIAO C B, et al. A lightweight YOLOv4-EDAM model for accurate and real-time detection of foreign objects suspended on power lines[J]. IEEE Transactions on Power Delivery, 2023, 38(2): 1329-1340.
[18]	WOO S, PARK J, LEE J Y, et al. CBAM: convolutional block attention module[M]//Computer Vision - ECCV 2018. Cham: Springer International Publishing, 2018: 3-19.
[19]	张焕龙, 齐企业, 张杰, 等. 基于改进YOLOv5的输电线路鸟巢检测方法研究[J]. 电力系统保护与控制, 2023, 51(2): 151-159.
	ZHANG H L, QI Q Y, ZHANG J, et al. Bird nest detection method for transmission lines based on improved YOLOv5[J]. Power System Protection and Control, 2023, 51(2): 151-159 (in Chinese).
[20]	宋智伟, 黄新波, 纪超, 等. 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.
	SONG Z W, HUANG X B, JI C, et al. Transmission Line Connection Fittings and Corrosion Detection Method Based on Pcsa-Yolov7 Former[J]. Electric Power, 2024, 57(6): 141-152 (in Chinese).
[21]	黄悦华, 刘恒冲, 陈庆, 等. 基于USRNet与改进YOLOv5x的输电线路绝缘子故障检测方法[J]. 高电压技术, 2022, 48(9): 3437-3446.
	HUANG Y H, LIU H C, CHEN Q, et al. Transmission line insulator fault detection method based on USRNet and improved YOLOv5x[J]. High Voltage Engineering, 2022, 48(9): 3437-3446 (in Chinese).
[22]	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
[23]	张烨, 李博涛, 尚景浩, 等. 基于多尺度卷积注意力机制的输电线路防振锤缺陷检测[J]. 电工技术学报, 2024, 39(11): 3522-3537.
	ZHANG Y, LI B T, SHANG J H, et al. Defect detection of transmission line damper based on multi-scale convolutional attention mechanism[J]. Transactions of China Electrotechnical Society, 2024, 39(11): 3522-3537 (in Chinese).
[24]	郑含博, 胡思佳, 梁炎燊, 等. 基于YOLO-2MCS的输电线路走廊隐患目标检测方法[J]. 电工技术学报, 2024, 39(13): 4164-4175.
	ZHENG H B, HU S J, LIANG Y S, et al. A hidden danger object detection method for transmission line corridor based on YOLO-2MCS[J]. Transactions of China Electrotechnical Society, 2024, 39(13): 4164-4175 (in Chinese).
[25]	郝帅, 张旭, 马旭, 等. 基于PKAMNet的输电线路小目标故障检测方法[J]. 高电压技术, 2023, 49(8): 3385-3394.
	HAO S, ZHANG X, MA X, et al. Small target fault detection method for transmission lines based on PKAMNet[J]. High Voltage Engineering, 2023, 49(8): 3385-3394 (in Chinese).
[26]	WANG Q L, WU B G, ZHU P F, et al. ECA-net: efficient channel attention for deep convolutional neural networks[C]// 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2020: 11531-11539.
[27]	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. New York: IEEE Press, 2021: 13708-13717.
[28]	OUYANG D L, HE S, ZHANG G Z, et al. Efficient multi-scale attention module with cross-spatial learning[C]// 2023 IEEE International Conference on Acoustics, Speech and Signal Processing. New York: IEEE Press, 2023: 1-5.
[29]	HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]// 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2018: 7132-7141.
[30]	ZHANG Q L, YANG Y B. SA-net: shuffle attention for deep convolutional neural networks[C]// 2021 IEEE International Conference on Acoustics, Speech and Signal Processing. New York: IEEE Press, 2021: 2235-2239.
[31]	LIU Y C, SHAO Z R, HOFFMANN N. Global attention mechanism: retain information to enhance channel-spatial interactions[EB/OL]. [2024-01-12]. http://arxiv.org/abs/2112.05561.

硬件名称	型号
CPU	Intel(R) Core(TM) i9-10850K
GPU	GeForce RTX 3080
操作系统	Ubuntu 18.04
框架	Pytorch 1.12.0
计算架构	Cuda 11.3
语言	Python 3.10

硬件名称	型号
CPU	Intel(R) Core(TM) i9-10850K
GPU	GeForce RTX 3080
操作系统	Ubuntu 18.04
框架	Pytorch 1.12.0
计算架构	Cuda 11.3
语言	Python 3.10

Method	P	R	AP50		mAP50	mAP50:95
Method	P	R	nest类	other类	mAP50	mAP50:95
YOLOv8s (基线模型)	78.4	57.8	69.5	63.5	66.5	33.0
YOLOv8s+MIX	84.9	61.8	71.9	69.1	70.5	35.1
YOLOv8s+DBD	89.3	61.4	70.3	69.1	69.7	34.2
Ours	91.3	62.9	73.0	73.3	73.2	35.8

Method	P	R	AP50		mAP50	mAP50:95
Method	P	R	nest类	other类	mAP50	mAP50:95
YOLOv8s (基线模型)	78.4	57.8	69.5	63.5	66.5	33.0
YOLOv8s+MIX	84.9	61.8	71.9	69.1	70.5	35.1
YOLOv8s+DBD	89.3	61.4	70.3	69.1	69.7	34.2
Ours	91.3	62.9	73.0	73.3	73.2	35.8

目标尺度	Method	P/%	R/%	AP50/%		mAP50/%	mAP50:95_/%
目标尺度	Method	P/%	R/%	nest类	other类	mAP50/%	mAP50:95_/%
小尺度	YOLOv8s	80.1	60.7	50.2	99.5	74.8	39.1
	YOLOv8s+MIX	85.7	70.2	54.1	99.5	76.8	40.7
	YOLOv8s+DBD	85.3	71.3	56.7	99.5	78.1	39.9
	Ours	86.1	70.5	61.2	99.5	80.3	42.3
中尺度	YOLOv8s	85.3	68.0	76.3	70.3	73.3	34.9
	YOLOv8s+MIX	85.9	65.4	80.4	73.2	76.8	35.8
	YOLOv8s+DBD	91.0	68.5	76.1	77.2	76.7	35.7
	Ours	95.0	68.6	76.5	75.1	75.8	34.2
大尺度	YOLOv8s	91.4	62.0	87.4	56.9	72.1	39.5
	YOLOv8s+MIX	89.9	65.2	86.7	64.9	75.8	42.3
	YOLOv8s+DBD	88.4	66.9	86.0	61.6	73.8	41.0
	Ours	91.4	65.8	87.5	69.5	78.5	43.5