Abstract:The defect detection algorithm for power transmission lines based on deep learning heavily relies on the dataset used. The learned features of line defects and image features are highly correlated with the feature distribution of the dataset used. Currently, most datasets used to train defect detection algorithms are captured in a single environment, leading to the specific defect recognition algorithm performing well only in a particular environment. However, its generalization ability is inadequate in different shooting environments, such as low light, haze, and blur, resulting in poor algorithm robustness. To improve the robustness of the intelligent defect detection algorithm for power transmission lines in different environmental conditions and overcome the issue of decreased recognition performance in different environments, this paper proposes the cross-environment robust YOLO algorithm (ER-YOLO). Firstly, based on the you only look once v8 (YOLOv8) algorithm, ER-YOLO enhances the algorithm's long-distance modeling capability by employing the Transformer attention mechanism based on the generalized attention theory. Secondly, ER-YOLO improves the path aggregation network (PANet) in the YOLOv8 algorithm by increasing the convolutional kernel size and introducing an efficient attention mechanism in the cross stage partial layer (CSPLayer), enhancing the network's object detection capability. Finally, ER-YOLO uses a multiple attention mechanism detection head network to strengthen the algorithm's multi-scale, spatial location, and multi-task perception capabilities, enabling the network to focus on crucial target information. Additionally, to validate the cross-environment detection performance of the defect recognition algorithm, a certain number of corresponding datasets of power transmission line defects in different environments are required. However, due to environmental and practical engineering constraints, collecting a large number of datasets in specific environments is challenging. To address the issue of the test dataset, this paper proposes common methods for generating adverse environment data for power transmission line defect recognition. High-fidelity test datasets were generated based on normal environment data. The paper explores methods for simulating adverse environments, including dark light environment simulation using exposure fusion algorithms and brightness reduction methods, haze environment simulation using Cycle GAN networks, and imaging blur environment simulation using mean filtering methods. The effectiveness of each method was evaluated and compared with other methods, providing conditions for testing the robustness of defect recognition algorithms across different environments. Through ablation experiments and comparative analysis, ER-YOLO demonstrated higher defect recognition accuracy and robustness in cross-environment testing. The average mAP value under various test datasets was 0.726, showing an improvement of 0.069 compared to the previous version. The algorithm's effectiveness was further validated in real environments. This study proposes a defect recognition method for power transmission lines across shooting environments, exhibiting excellent performance in cross-environment recognition. It also explores methods for generating cross-environment images, providing insights for future virtual dataset generation techniques. Future research directions may focus on cross-environment defect recognition studies for other types of defects and explore other effective methods for generating multi-environment virtual datasets.
裴少通, 张行远, 胡晨龙, 杨文杰, 刘云鹏. 基于ER-YOLO算法的跨环境输电线路缺陷识别方法[J]. 电工技术学报, 2024, 39(9): 2825-2840.
Pei Shaotong, Zhang Hangyuan, Hu Chenlong, Yang Wenjie, Liu Yunpeng. The Defect Detection Method for Cross-Environment Power Transmission Line Based on ER-YOLO Algorithm. Transactions of China Electrotechnical Society, 2024, 39(9): 2825-2840.
[1] Nguyen V N, Jenssen R, Roverso D.Automatic autonomous vision-based power line inspection: a review of current status and the potential role of deep learning[J]. International Journal of Electrical Power & Energy Systems, 2018, 99: 107-120. [2] 胡晨龙, 裴少通, 刘云鹏, 等. 基于LEE-YOLOv7的输电线路边缘端实时缺陷检测方法[J/OL]. 高电压技术, 2023: 1-14[2023-12-09]. https://doi.org/10.13336/j.1003-6520.hve.20230945. Hu Chenlong, Pei Shaotong, Liu Yunpeng, et al. Real-time defect detection method for transmission line edge end based on LEE-YOLOv7[J/OL]. High Voltage Engineering, 2023: 1-14[2023-12-09]. https://doi.org/10.13336/j.1003-6520.hve.20230945. [3] 周宇, 徐波, 宋爱国, 等. 基于改进文本检测识别的绝缘子串异常定位和判别方法[J]. 高电压技术, 2021, 47(11): 3819-3826. Zhou Yu, Xu Bo, Song Aiguo, et al.Anomaly location and discrimination method of insulator string based on improved text detection and recognition[J]. High Voltage Engineering, 2021, 47(11): 3819-3826. [4] 李斌, 屈璐瑶, 朱新山, 等. 基于多尺度特征融合的绝缘子缺陷检测[J]. 电工技术学报, 2023, 38(1): 60-70. Li Bin, Qu Luyao, Zhu Xinshan, et al.Insulator defect detection based on multi-scale feature fusion[J]. Transactions of China Electrotechnical Society, 2023, 38(1): 60-70. [5] 苟军年, 杜愫愫, 刘力. 基于改进掩膜区域卷积神经网络的输电线路绝缘子自爆检测[J]. 电工技术学报, 2023, 38(1): 47-59. Gou Junnian, Du Susu, Liu Li.Transmission line insulator self-explosion detection based on improved mask region-convolutional neural network[J]. Tran-sactions of China Electrotechnical Society, 2023, 38(1): 47-59. [6] 宋立业, 刘帅, 王凯, 等. 基于改进EfficientDet的电网元件及缺陷识别方法[J]. 电工技术学报, 2022, 37(9): 2241-2251. Song Liye, Liu Shuai, Wang Kai, et al.Identification method of power grid components and defects based on improved EfficientDet[J]. Transactions of China Electrotechnical Society, 2022, 37(9): 2241-2251. [7] 仲林林, 胡霞, 刘柯妤. 基于改进生成对抗网络的无人机电力杆塔巡检图像异常检测[J]. 电工技术学报, 2022, 37(9): 2230-2240, 2262. Zhong Linlin, Hu Xia, Liu Keyu.Power tower anomaly detection from unmanned aerial vehicles inspection images based on improved generative adversarial network[J]. Transactions of China Electrotechnical Society, 2022, 37(9): 2230-2240, 2262. [8] 游越, 伊力哈木·亚尔买买提. 基于改进YOLOv5在电力巡检中的目标检测算法研究[J]. 高压电器, 2023, 59(2): 89-96. You Yue, Yilihamu Yaermaimaiti.Research on target detection algorithm based on improved YOLOv5 in power partrol inspection[J]. High Voltage Apparatus, 2023, 59(2): 89-96. [9] 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/OL]. 电工技术学报, 2023: 1-12[2023-12-09]. https://doi.org/10.19595/j.cnki.1000-6753.tces.230666. Zheng Hanbo, Hu Sijiao, Liang Yanshen, et al. A hidden danger object detection method for trans-mission linecorridor based on YOLO-2MCS[J/OL]. Transactions of China Electrotechnical Society, 2023: 1-12[2023-12-09]. https://doi.org/10.19595/j.cnki.1000-6753.tces.230666. [10] 张芯睿, 赵清华, 王雷, 等. 基于Mask R-CNN的雾天场景目标检测[J]. 电光与控制, 2022, 29(12): 83-88. Zhang Xinrui, Zhao Qinghua, Wang Lei, et al.Target detection in foggy scene based on Mask R-CNN[J]. Electronics Optics & Control, 2022, 29(12): 83-88. [11] 郭克友, 王苏东, 李雪, 等. 基于Dim env-YOLO算法的昏暗场景车辆多目标检测[J]. 计算机工程, 2023, 49(3): 312-320. Guo Keyou, Wang Sudong, Li Xue, et al.Multi-target detection of vehicles in dim scenes based on Dim env-YOLO algorithm[J]. Computer Engineering, 2023, 49(3): 312-320. [12] 翟永杰, 杨珂, 王乾铭, 等. 基于混合样本迁移学习的盘型绝缘子缺陷检测[J]. 中国电机工程学报, 2023, 43(7): 2867-2876. Zhai Yongjie, Yang Ke, Wang Qianming, et al.Disc insulator defect detection based on mixed sample transfer learning[J]. Proceedings of the CSEE, 2023, 43(7): 2867-2876. [13] 刘庆臻, 刘亚东, 严英杰, 等. 基于域随机化的绝缘子缺损数据自动生成与评价方法[J/OL]. 高电压技术, 2023: 1-13[2023-12-09]. DOI:10.13336/j.1003-6520.hve.20231241. Liu QingZhen, Liu Yadong, Yan jieing, et al. Automatic generation and evaluation method of insulator broken defect data based on domain randomization[J/OL]. High Voltage Engineering, 2023: 1-13[2023-12-09]. DOI:10.13336/j.1003-6520.hve.20231241. [14] Redmon J, Divvala S, Girshick R, et al.You only look once: unified, real-time object detection[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, 2016: 779-788. [15] Feng Chengjian, Zhong Yujie, Gao Yu, et al.TOOD: task-aligned one-stage object detection[C]//2021 IEEE/CVF International Conference on Computer Vision (ICCV), Montreal, QC, Canada, 2021: 3490-3499. [16] Zheng Zhaohui, Wang Ping, Ren Dongwei, et al.Enhancing geometric factors in model learning and inference for object detection and instance segmentation[J]. IEEE Transactions on Cybernetics, 2022, 52(8): 8574-8586. [17] 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, CA, USA, 2017: 6000-6010. [18] Carion N, Massa F, Synnaeve G, et al.End-to-end object detection with transformers[C]//European Conference on Computer Vision - ECCV 2020, Glasgow, UK, 2020: 213-229. [19] Dai Zihang, Yang Zhilin, Yang Yiming, et al.Transformer-XL: attentive language models beyond a fixed-length context[C]//Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, Florence, Italy, 2019: 2978-2988. [20] Zhu Xizhou, Cheng Dazhi, Zhang Zheng, et al.An empirical study of spatial attention mechanisms in deep networks[C]//2019 IEEE/CVF International Conference on Computer Vision (ICCV), Seoul, Korea (South), 2019: 6687-6696. [21] Liu Zhuang, Mao Hanzi, Wu Chaoyuan, et al.A ConvNet for the 2020s[C]//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), New Orleans, LA, USA, 2022: 11966-11976. [22] Ding Xiaohan, Zhang Xiangyu, Han Jungong, et al.Scaling up your kernels to 31x31: revisiting large kernel design in CNNs[C]//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), New Orleans, LA, USA, 2022: 11953-11965. [23] Lü Chengqi, Zhang Wenwei, Huang Haian, et al. RTMDet: an empirical study of designing real-time object detectors[J/OL]. ArXiv, 2022: 1-15[2023-12-09]. https://doi.org/10.48550/arXiv.2212.07784. [24] Chollet F.Xception: deep learning with depthwise separable convolutions[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, USA, 2017: 1800-1807. [25] Wang Qilong, Wu Banggu, Zhu Pengfei, et al.ECA-Net: efficient channel attention for deep convolutional neural networks[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA, 2020: 11531-11539. [26] Dai Xiyang, Chen Yinpeng, Xiao Bin, et al.Dynamic head: unifying object detection heads with attentions[C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Nashville, TN, USA, 2021: 7369-7378. [27] Chen Yinpeng, Dai Xiyang, Liu Mengchen, et al.Dynamic ReLU[C]//European Conference on Computer Vision - ECCV 2020, Glasgow, UK, 2020: 351-367. [28] Ying Zhenqiang, Li Ge, Gao Wen. A bio-inspired multi-exposure fusion framework for low-light image enhancement[J/OL]. ArXiv, 2017: 1-10[2023-12-09]. https://doi.org/10.48550/arXiv.1711.00591. [29] Goodfellow I, Pouget-Abadie, Mirza M, et al. Generative adversarial networks[J/OL]. ArXiv, 2014: 1406.2661vl. https://doi.org/10.48550/arXiv.1406.2661. [30] Zhu Junyan, Park T, Isola P, et al.Unpaired image-to-image translation using cycle-consistent adversarial networks[C]//2017 IEEE International Conference on Computer Vision (ICCV), Venice, Italy, 2017: 2242-2251. [31] 赵书涛, 王紫薇, 陈志华, 等. 有载分接开关GLCM纹理特征及改进随机森林算法的故障诊断方法[J]. 高电压技术, 2022, 48(9): 3593-3601. Zhao Shutao, Wang Ziwei, Chen Zhihua, et al.GLCM texture features of on-load tap changer and fault diagnosis method based on im-proved random forest algorithm[J]. High Voltage Engineering, 2022, 48(9): 3593-3601. [32] Chen Kai, Wang Jiaqi, Pang Jiangmiao, et al.MMDetection: open MMLab detection toolbox and benchmark[J/OL]. ArXiv, 2019: 1-13[2023-12-09]. https://doi.org/10.48550/arXiv.1906.07155. [33] MMYOLO: OpenMMLab YOLO series toolbox and benchmark[EB/OL].(2023-08-15)[2023-12-09]. https://github.com/open-mmlab/mmyolo. [34] Tian Zhi, Shen Chunhua, Chen Hao, et al.FCOS: fully convolutional one-stage object detection[C]//2019 IEEE/CVF International Conference on Computer Vision (ICCV), Seoul, Korea (South), 2019: 9626-9635. [35] Kim K, Lee H S.Probabilistic anchor assignment with IoU prediction for object detection[C]//European Conference on Computer Vision-ECCV 2020, Glasgow, UK, 2020: 355-371.