融合高效注意力的多尺度输电线路部件检测

doi:10.3778/j.issn.1002-8331.2209-0125

摘要/Abstract

摘要： 针对在高分辨率输电线路图像中，不同种类部件尺度跨越大，难以被均衡检测的问题，提出一种融合高效注意力的多尺度输电线路部件检测算法。在YOLO v5目标检测算法中，设计添加高效注意力模块ECBAM提高算法特征提取能力。根据输电线路部件的特征分布统计，使用滑动窗口对高分辨率输电线路图像进行切片，并对切片前后的图像分别使用改进后的YOLO v5算法训练模型。将两个模型的检测结果进行集成，得到多尺度输电线路部件检测结果。在公开的PLAD架空输电线路图像数据集上，该模型的检测性能远超现有目标检测模型，Precision可达83.2%，Recall可达92.8%，相比数据集原作者提出的模型，mAP值提升了1.6个百分点，达到了90.8%，且能检测出未在原始数据集上标注出的隐蔽目标，验证了在高分辨率图像中检测多尺度输电线路部件的有效性。

关键词: 输电线路, 多尺度目标检测, 滑窗切片, 注意力机制

Abstract: For the problem that different types of components span large scales and are difficult to be detected in a balanced manner in high-resolution transmission line images, a multi-scale transmission line component detection algorithm incorporating efficient attention is proposed. Firstly, an efficient attention module ECBAM is designed and added to the YOLO v5 object detection algorithm to improve the feature extraction capability of the algorithm. Secondly, according to the feature distribution statistics of transmission line components, the high-resolution transmission line images are sliced by sliding window, and the improved YOLO v5 algorithm is used separately to train the models for the images before and after slicing. Finally, the detection results of the two models are integrated to obtain the detection results of multi-scale transmission line components. On the publicly available PLAD overhead transmission line image dataset, the detection performance of the proposed model far exceeds existing object detection models, with Precision up to 83.2% and Recall up to 92.8%. Compared with the model proposed by the original authors of the dataset, the mAP value improves by 1.6 percentage points to 90.8% and it can detect hidden objects that are not labeled on the original dataset, which verifies the effectiveness of detecting multi-scale transmission line components in high-resolution images.

Key words: transmission line, multi-scale object detection, sliding window slicing, attention mechanism

陈思雨, 付章杰. 融合高效注意力的多尺度输电线路部件检测[J]. 计算机工程与应用, 2024, 60(1): 327-336.

CHEN Siyu, FU Zhangjie. Multi-Scale Transmission Line Component Detection Incorporating Efficient Attention[J]. Computer Engineering and Applications, 2024, 60(1): 327-336.

参考文献

[1] 周帅. 无人机在输电线路巡检中的应用[J]. 电子技术, 2021, 50(8): 276-277.
ZHOU S. Application of UAVs in the inspection of power transmission lines[J]. Electronic Technology, 2021, 50(8): 276-277.
[2] LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot multibox detector[C]//Proceedings of the 2016 European Conference on Computer Vision. Cham: Springer, 2016: 21-37.
[3] LIN T, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection[C]//Proceedings of the 2017 IEEE International Conference on Computer Vision. Piscataway: IEEE, 2017: 2980-2988.
[4] REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection[C]//Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2016: 779-788.
[5] REDMON J, FARHADI A. YOLO9000: better, faster, stronger[C]//Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2017: 7263-7271.
[6] REDMON J, FARHADI A. Yolov3: an incremental improvement[EB/OL]. (2018-04-08)[2022-05-20]. https://arxiv.org/pdf/1804.02767.pdf.
[7] BOCHKOVSKIY A, WANG C, LIAO H M. Yolov4: optimal speed and accuracy of object detection[EB/OL]. (2020-04-23)[2022-05-20]. https://arxiv.org/pdf/2004.10934.pdf.
[8] Ultralytics. YOLO v5[CP/OL]. https://github.com/ultralytics/yolov5.
[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, 2015, 38(1): 142-158.
[10] GIRSHICK R. Fast R-CNN[C]//Proceedings of the 2015 IEEE International Conference on Computer Vision. Piscata-way: IEEE, 2015: 1440-1448.
[11] REN S, HE K, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[C]//Advances in Neural Information Processing Systems, 2015: 91-99.
[12] 王万国, 田兵, 刘越, 等. 基于RCNN的无人机巡检图像电力小部件识别研究[J]. 地球信息科学学报, 2017, 19(2): 256-263.
WANG W G, TIAN B, LIU Y, et al. Research on power widget recognition of UAV inspection images based on RCNN[J]. Journal of Geoinformation Science, 2017, 19(2): 256-263.
[13] 戚银城, 江爱雪, 赵振兵, 等. 基于改进SSD模型的输电线路巡检图像金具检测方法[J]. 电测与仪表, 2019, 56(22): 7-12.
QI Y C, JIANG A X, ZHAO Z B, et al. Improved SSD model based on transmission line inspection image metallic detection method[J]. Electrical Measurement and Instrumentation, 2019, 56(22): 7-12.
[14] 郭敬东, 陈彬, 王仁书, 等. 基于YOLO的无人机电力线路杆塔巡检图像实时检测[J]. 中国电力, 2019, 52(7): 17-23.
GUO J D, CHEN B, WANG R S, et al. Real-time detection of UAV power line pole tower inspection images based on YOLO[J]. China Electric Power, 2019, 52(7): 17-23.
[15] 张鸥, 徐强胜, 刘靖波, 等. 无人机巡检图像电力小部件识别技术研究[J]. 科技创新导报, 2019, 16(14): 110-112.
ZHANG O, XU Q S, LIU J B, et al. Research on recognition technology of small parts of electric power from drone inspection images[J]. Science and Technology Innovation Herald, 2019, 16(14): 110-112.
[16] 赵振兵, 李延旭, 戚银城, 等. 基于动态焦点损失函数和样本平衡方法的绝缘子缺陷检测方法[J]. 电力自动化设备, 2020, 40(10): 205-211.
ZHAO Z B, LI Y X, QI Y C, et al. Insulator defect detection method based on dynamic focal loss function and sample balancing method[J]. Power Automation Equipment, 2020, 40(10): 205-211.
[17] 韩汉贤, 罗金满, 刘丽媛, 等. 基于无人机图像的输电线路部件检测方法研究[J/OL]. 电测与仪表: 1-7[2022-07-27].http://kns.cnki.net/kcms/detail/23.1202.TH.20220606.1504.
009.html.
HAN H X, LUO J M, LIU L Y, et al. Research on the detection method of transmission line components based on UAV images[J/OL]. Electrical Measurement and Instrumentation: 1-7[2022-07-27]. http://kns.cnki.net/kcms/detail/23.
1202.TH.20220606.1504.009.html.
[18] 董召杰. 基于YOLOv3的电力线关键部件实时检测[J]. 电子测量技术, 2019, 42(23): 173-178.
DONG Z J. YOLOv3-based real-time inspection of key components in power lines[J]. Electronic Measurement Technology, 2019, 42(23): 173-178.
[19] ZHAI Y, WANG Q, YANG X, et al. Multi-fitting detection on transmission line based on cascade reasoning graph network[J]. IEEE Transactions on Power Delivery, 2022, 37(6): 4858-4868.
[20] WOO S, PARK J, LEE J, et al. CBAM: convolutional block attention module[C]//Proceedings of the 2018 European Conference on Computer Vision. Cham: Springer, 2018: 3-19.
[21] WANG Q, WU B, ZHU P, et al. ECA-Net: efficient channel attention for deep convolutional neural networks[C]//Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 13-19.
[22] 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]//Proceedings of the 2021 34th SIBGRAPI Conference on Graphics, Patterns and Images. Piscataway: IEEE, 2021: 215-222.