State Recognition Method of Transmission Line Ice-Melting Switch Based on Improved YOLOv7

doi:10.3778/j.issn.1002-8331.2307-0068

Abstract

Abstract: Automatic identification of the state of disconnecting switch is deemed crucial for the intelligent ice-melting reversal operation of transmission lines during ice periods. Acknowledging the low accuracy of traditional image recognition methods when identifying ice-melting switches under severe weather conditions, a method predicated on an improved YOLOv7 framework for recognizing the state of ice-melting switches is proposed. A self-attention (S-A) mechanism is integrated into the YOLOv7 network to enhance global feature extraction capability in low-contrast images. Concurrently, the SPPCSPC module in the network is modified with the inclusion of the atrous spatial pyramid pooling (ASPP) technique, aimed at augmenting the recognition capability for larger targets such as the overlapped switch. Given the distinct structure, size, and position of the overlapped switch, constraint terms are introduced to the loss function to enhance its recognition specificity. An M-MBO acceleration network is subsequently designed, utilizing a multi-branch architecture to simplify the model during inference, thereby improving recognition speed. Experimental results reveals that the refined YOLOv7 model achieves a mean average precision (mAP) value of 97.9%, marking an increase of 2.5 percentage points in average precision compared to previous methods, underscoring the effectiveness of the proposed approach.

Key words: YOLOv7, switch state recognition, self-attention mechanism, atrous spatial pyramid pooling (ASPP), loss function constraint term, M-MBO

摘要： 隔离刀闸状态的自动识别是冰期输电线路智能融冰倒闸操作中的关键环节。针对恶劣天气条件下，传统图像识别方法在识别融冰刀闸时精度较低的问题，提出了一种基于改进YOLOv7的融冰刀闸状态识别方法。在YOLOv7网络中引入自注意力机制（self-attention，S-A）模块，以增强网络在低对比度图像中的全局特征提取能力。同时对网络中的SPPCSPC模块进行改进，引入空洞空间金字塔池化技术（atrous spatial pyramid pooling，ASPP），提高对搭接刀闸等此类大目标的识别能力。根据搭接刀闸的特殊结构、大小和位置，在损失函数中添加约束项，增强对刀闸识别的针对性。最后，设计了一个M-MBO加速网络，利用多分支架构在推理时简化模型，提高模型识别速度。实验结果表明，在保证识别速度的同时，改进的YOLOv7模型mAP值可达97.9%，相比改进前的方法平均精度均值提高了2.5个百分点，验证了该方法的有效性。

关键词: YOLOv7, 刀闸状态识别, 自注意力机制, 空洞空间金字塔池化（ASPP）, 损失函数约束项, M-MBO

GAO Xujie, LI Zetao, ZENG Huarong, YANG Qi, ZHANG Lusong. State Recognition Method of Transmission Line Ice-Melting Switch Based on Improved YOLOv7[J]. Computer Engineering and Applications, 2024, 60(23): 314-324.

高绪杰, 李泽滔, 曾华荣, 杨旗, 张露松. 改进YOLOv7的输电线路融冰刀闸状态识别方法[J]. 计算机工程与应用, 2024, 60(23): 314-324.

References

[1] 张浩, 王玮, 徐丽杰, 等. 图像识别技术在电力设备监测中的应用[J]. 电力系统保护与控制, 2010, 38(6): 88-91.
ZHANG H, WANG W, XU L J, et al. Application of image recognition technology in power equipment monitoring[J]. Power System Protection and Control, 2010, 38(6): 88-91.
[2] 丁四海, 刘玉雪, 路林吉. 数字图像处理技术在电气控制柜开关状态识别中的应用[J]. 微型电脑应用, 2013, 30(5): 39-40.
DING S H, LIU Y X, LU L J. Application of digital image processing technology in electrical control cabinet switch state identification[J]. Microcomputer Applications, 2013, 30(5): 39-40.
[3] LIU Y, YONG J, LIU L, et al. The method of insulator recognition based on deep learning[C]//Proceedings of the International Conference on Applied Robotics for the Power Industry, 2016: 1-5.
[4] LIN H, ZHANG W, et al. A condition monitoring algorithm based on image geometric analysis for substation switch[C]//Proceedings of the International Conference on Intelligent Computing and Internet of Things, 2015: 72-76.
[5] 宋亚凯, 樊晓虹, 何国锋, 等. 基于轮廓匹配度的隔离开关三相分合闸状态同步识别算法研究[J]. 高压电器, 2022, 58(4): 151-157.
SONG Y K, FAN X H, HE G F, et al. Synchronous recognition algorithm of three-phase opening and closing state of disconnector based on contour matching degree[J]. High Voltage Apparatus, 2022, 58(4): 151-157.
[6] 邵剑雄, 闫云凤, 齐冬莲. 基于霍夫森林的变电站开关设备检测及状态识别[J]. 电力系统自动化, 2016, 40(11): 115-120.
SHAO J X, YAN Y F, QI D L. Substation switch detection and state recognition based on Hough forests[J]. Power System Automation, 2016, 40(11): 115-120.
[7] KHAN D. Deep learning based power switch detection and state recognition[D]. Chengdu: University of Electronic Science and Technology of China, 2018.
[8] 张骥, 张金锋, 朱能富, 等. 基于改进深度学习的刀闸状态识别方法研究[J]. 电测与仪表, 2018, 55(5): 8-13.
ZHANG J, ZHANG J F, ZHU N F, et al. Research of the switch state recognition method based on the improved deep learning[J]. Electrical Measurement & Instrumentation, 2018, 55(5): 8-13.
[9] CHEN T, ZHANG G Y, QI D L. A recognition method of smart substation switchgear state based on fully convolutional networks[C]//Proceedings of the 35th Chinese Control Conference, 2016: 9894-9897.
[10] 刘梓权, 王慧芳, 管敏渊, 等. 隔离开关图像数据扩充方法及其在自动状态识别中的应用[J]. 高电压技术, 2020, 46(2): 441-447.
LIU Z Q, WANG H F, GUAN M Y, et al. Isolation switch image data expansion method and it’s application in automatic state identification[J]. High Voltage Technology, 2020, 46(2): 441-447.
[11] 尤振飞, 赵健, 王小宇, 等. 基于语义分割与连通区域标记的隔离开关状态识别方法[J]. 电力系统自动化, 2021, 45(20): 157-165.
YOU Z F, ZHAO J, WANG S Y, et al. State recognition method for disconnector based on semantic segmentation and connected component labeling[J]. Power System Automation, 2021, 45(20): 157-165.
[12] 李文亮. 雷达点云与图像结合的刀闸状态精确检测研究[D]. 广州: 华南理工大学, 2022.
LI W L. Research on accurate detection of the stateof isolator based on radar point cloud and image[D]. Guangzhou: South China University of Technology, 2022.
[13] REDMON J, FARHADI A. YOLOv3: an incrementalimprovement[J]. arXiv:1804.02767, 2018.
[14] BOCHKOVSKIY A, WANG C Y, LIAO H Y M. YOLO-v4: optimal speed and accuracy of object detection[J]. arXiv: 2004.10934, 2020.
[15] 王鹏, 王玉林, 焦博文, 等. 基于YOLOv5的道路目标检测算法研究[J]. 计算机工程与应用, 2023, 59(1): 117-125.
WANG P, WANG Y L, JIAO B W, et al. Research on road target detection algorithm based on YOLOv5[J]. Computer Engineering and Applications, 2023, 59(1): 117-125.
[16] WANG C Y, BOCHKOVSKIY A, LIAO H. YOLOv7: trainable bag-of-freebies sets new state-of-the-art for realtime object detectors[J]. arXiv:2207.02696, 2022.
[17] WANG X, GIRSHICK R, GUPTA A, et al. Non-local neural networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018: 7794-7803.
[18] CHEN L C, PAPANDREOU G, KOK-KINOS I, et al. DeepLab: semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected CRFs[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 40(4): 834-848.
[19] HE K, ZHANG X Y, REN S, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(9): 1904-1916.
[20] LIN T Y, DOLLAR P, GIRSHICK R, et al. Feature pyramid networks for object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017: 2117-2125.
[21] LIU S, QI L, QIN H F, et al. Path aggregation net-work for instance segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018: 8759-8768.
[22] VASU P K A, GABRIEL J, ZHU J , et al. MobileOne: an improved one millisecond mobile backbone[C]//Proceedings of the 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023: 7907-7917.
[23] POWERS D M W. Evaluation: from precision, recall and F-measure to ROC, informedness, markedness＆correlation[J].arXiv:2010.16061, 2020.
[24] ZHENG Z, WANG P, LIU W, et al. Distance-IoU l-oss: faster and better learning for bounding box regression[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2020: 12993-13000.