Weak supervision detection method for pin-missing bolts of transmission lines based on SAW-PCL
-
摘要:
螺栓作为输电线路中不可或缺的紧固件，其缺销必然会引起重大的安全隐患。针对螺栓目标较小、标注难度大的问题，提出了一种基于SAW-PCL的输电线路缺销螺栓弱监督检测方法。该方法通过图像级标注信息即可定位到螺栓目标。在主网络中引入卷积块注意模块(CBAM)，抑制无用的背景特征，提取螺栓精细特征，提高螺栓的检测能力。针对弱监督检测中缺销螺栓的检测精度远低于正常螺栓及不平衡性问题，提出自适应加权损失函数(SAW)，动态调节模型对不同类别样本的学习程度，均衡不同类别之间的检测精度，并定义了平均类间检测精度差(ADPD)来评价不平衡性。构建的自适应加权损失函数可以提升缺销螺栓的检测精度，对正常螺栓和缺销螺栓的检测精度有一定的均衡能力，定义的ADPD可以评价模型检测性能的平衡度。在自建数据集V1上的实验结果表明：改进方法的平均准确率均值(mAP)提高了19.7%，ADPD值降低了21.8，在mAP和ADPD双重指标评估下的模型表现出了更好的缺销螺栓检测能力。
Abstract:Bolt is an indispensable fastener in the transmission line, and a pin-missing bolt will inevitably cause major safety hazards. Since the bolt target is small, and the annotation is difficult, a weak supervision detection method for pin-missing bolts of transmission lines based on SAW-PCL was proposed, and the bolt target could be located through image-level annotation information. The convolutional block attention module (CBAM) was introduced into the main network to suppress useless background features, extract fine features of bolts, and improve the detection capability of bolts. In view of the imbalance problem that the detection accuracy of the pin-missing bolt was far lower than that of the normal bolt in the weak supervision detection, an self-adaptation weighted loss function (SAW) was proposed to dynamically adjust the learning degree of the model for different categories of samples, balance the detection accuracy between different categories, and focus on the problem of pin-missing bolts. Moreover, the average detection precision difference among classes (ADPD) was defined to evaluate this imbalance. The constructed SAW could improve the detection accuracy of pin-missing bolts and had a certain ability to balance the detection accuracy of normal bolts and pin-missing bolts. The defined average detection precision difference among classes could be used to evaluate the balance of the detection performance of the model. The experimental results on the self-built dataset V1 show that the mean average precision (mAP) of the improved algorithm is increased by 19.7%, and the ADPD value is reduced by 21.8. The model under the evaluation of indexes mAP and ADPD shows better detection ability of pin-missing bolts.
-
图 7 本文方法与PCL模型的螺栓图像检测效果对比
Figure 7. Comparison of bolt image detection effects based on the proposed method and PCL model
表 1 螺栓缺陷检测数据集V1
Table 1. Dataset V1 for pin-missing bolt detection
类别 训练集图像数量 测试集图像数量 正常螺栓 1089 274 缺销螺栓 1736 450 
下载: 导出CSV
表 2 螺栓缺陷检测数据集V2
Table 2. Dataset V2 for pin-missing bolt detection
类别 训练集图像数量 测试集图像数量 正常螺栓 1089 274 可见销子缺失螺栓 657 170 不可见销子缺失螺栓 1079 280
下载: 导出CSV
表 3 不同β 取值在测试集上mAP和ADPD对比
Table 3. Comparison of mAP and ADPD on test set with different β values
β AP/% mAP/% ADPD 正常螺栓 缺销螺栓 5 37.8 11.2 24.5 26.6 4 40.1 11.2 25.7 28.9 3 31.5 10.1 20.8 21.4 2 20.8 9.0 14.9 11.8 1 34.3 24.2 29.2 10.1
下载: 导出CSV
表 4 不同损失函数在测试集上mAP和ADPD对比
Table 4. Comparison of mAP and ADPD of different loss functions on test set
方法 AP/% mAP/% ADPD 正常螺栓 缺销螺栓 PCL模型 33.3 6.0 19.6 27.3 PCL模型+
加权损失函数49.2 16.1 32.7 33.1 SAW-PCL模型 34.3 24.2 29.2 10.1
下载: 导出CSV
表 5 在数据集V2上mAP和ADPD对比
Table 5. Comparison of mAP and ADPD on dataset V2
方法 AP/% mAP/% ADPD 正常螺栓 可见销子
缺失螺栓不可见销子
缺失螺栓PCL模型 24.3 6.3 18.2 16.3 12.0 SAW-PCL模型 35.8 28.2 21.6 28.5 9.5
下载: 导出CSV
表 6 在模型上添加CBAM模块进行的实验结果对比
Table 6. Comparison of experimental results by adding CBAM to model
方法 AP/% mAP/% ADPD 正常螺栓 缺销螺栓 PCL模型+CBAM 39.8 7.8 23.8 32.0 SAW-PCL模型+CBAM 42.0 36.5 39.3 5.5
下载: 导出CSV
-
[1] 雷朝煜, 郝良收, 戴甲水, 等. 高压直流输电线路故障定位研究综述[J]. 电力系统保护与控制, 2022, 50(11): 178-187.LEI Z Y, HAO L S, DAI J S, et al. A review of fault location methods in HVDC transmission lines[J]. Power System Protection and Control, 2022, 50(11): 178-187(in Chinese). [2] 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): 7423-7445.LIU C Y, WU Y Q. Research progress of vision detection methods based on deep learning for transmission lines[J]. Proceedings of the CSEE, 2023, 43(19): 7423-7445(in Chinese). [3] 蒲良, 张学军. 基于深度学习的无人机视觉目标检测与跟踪[J]. 北京航空航天大学学报, 2022, 48(5): 872-880.PU L, ZHANG X J. Deep learning based UAV vision object detection and tracking[J]. Journal of Beijing University of Aeronautics and Astronautics, 2022, 48(5): 872-880(in Chinese). [4] 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: 10.1109/TPAMI.2016.2577031 [5] REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: Unified, real-time object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2016: 779-788. [6] REDMON J, FARHADI A. YOLO9000: Better, faster, stronger[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2017: 6517-6525. [7] LIU W, ANGUELOV D, ERHAN D, et al. SSD: Single shot MultiBox detector[C]//Proceedings of the European Conference on Computer Vision. Berlin: Springer, 2016: 21-37. [8] TA Q, HUYNH T, QUANG Q, et al. Corrosion and bolt-loosening monitoring using image-based deep learning and Hough transform[C]//Proceedings of the 2020 World Congress on Advances in Civil, Environmental, and Materials Research and the 2020 Structures Congress. Seoul: IASEM, 2020. [9] GIRSHICK R, DONAHUE J, DARRELL T, et al. Region-based convolutional networks for accurate object detection and segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2016, 38(1): 142-158. doi: 10.1109/TPAMI.2015.2437384 [10] 薛阳, 吴海东, 张宁, 等. 基于改进Faster R-CNN输电线穿刺线夹及螺栓的检测[J]. 激光与光电子学进展, 2020, 57(8): 081008.XUE Y, WU H D, ZHANG N, et al. Detection of insulation piercing connectors and bolts on the transmission line using improved Faster R-CNN[J]. Laser & Optoelectronics Progress, 2020, 57(8): 081008(in Chinese). [11] 罗隆福, 叶威, 王健. 基于深度学习的高铁接触网顶紧螺栓的缺陷检测[J]. 铁道科学与工程学报, 2021, 18(3): 605-614.LUO L F, YE W, WANG J. Defect detection of the puller bolt in high-speed railway catenary based on deep learning[J]. Journal of Railway Science and Engineering, 2021, 18(3): 605-614(in Chinese). [12] WANG L M, TONG Z, JI B, et al. TDN: Temporal difference networks for efficient action recognition[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2021: 1895-1904. [13] 赵振兵, 金超熊, 戚银城, 等. 基于动态监督知识蒸馏的输电线路螺栓缺陷图像分类[J]. 高电压技术, 2021, 47(2): 406-414.ZHAO Z B, JIN C X, QI Y C, et al. Image classification of transmission line bolt defects based on dynamic supervision knowledge distillation[J]. High Voltage Engineering, 2021, 47(2): 406-414(in Chinese). [14] XIAO L, WU B, HU Y M. Missing small fastener detection using deep learning[J]. IEEE Transactions on Instrumentation and Measurement, 2020, 70: 3502209. [15] KIM Y. Convolutional neural networks for sentence classification[C]//Proceedings of the Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2014: 1746-1751. [16] HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2016: 770-778. [17] ZHAO Z B, QI H Y, QI Y C, et al. Detection method based on automatic visual shape clustering for pin-missing defect in transmission lines[J]. IEEE Transactions on Instrumentation and Measurement, 2020, 69: 6080-6091. doi: 10.1109/TIM.2020.2969057 [18] 李雪峰, 刘海莹, 刘高华, 等. 基于深度学习的输电线路销钉缺陷检测[J]. 电网技术, 2021, 45(8): 2988-2995.LI X F, LIU H Y, LIU G H, et al. Transmission line pin defect detection based on deep learning[J]. Power System Technology, 2021, 45(8): 2988-2995(in Chinese). [19] SHEN Y H, JI R R, WANG Y, et al. Enabling deep residual networks for weakly supervised object detection[C]//Proceedings of the European Conference on Computer Vision. Berlin: Springer, 2020: 118-136. [20] BILEN H, VEDALDI A. Weakly supervised deep detection networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2016: 2846-2854. [21] RUSSAKOVSKY O, DENG J, SU H, et al. ImageNet large scale visual recognition challenge[J]. International Journal of Computer Vision, 2015, 115(3): 211-252. doi: 10.1007/s11263-015-0816-y [22] TANG P, WANG X G, BAI X, et al. Multiple instance detection network with online instance classifier refinement[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2017: 3059-3067. [23] TANG P, WANG X G, BAI S, et al. PCL: Proposal cluster learning for weakly supervised object detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(1): 176-191. doi: 10.1109/TPAMI.2018.2876304 [24] WOO S, PARK J, LEE J Y, et al. CBAM: Convolutional block attention module[C]//Proceedings of the European Conference on Computer Vision. Berlin: Springer, 2018: 3-19. -
下载:
点击查看大图
计量
- 文章访问数: 240
- HTML全文浏览量: 104
- PDF下载量: 4
- 被引次数: 0
