基于通道剪枝的YOLOv7-tiny输电线路异物检测算法

doi:10.3778/j.issn.1002-8331.2311-0160

摘要/Abstract

摘要： 针对输电线路异物检测精度不佳且模型庞大的问题，提出了基于通道剪枝的改进YOLOv7-tiny算法用于输电线路异物检测。用ReXNet网络替代了YOLOv7-tiny的骨干网络改进原网络的特征瓶颈问题。引入了多样化分支块从而增加网络特征融合能力，通过基于层自适应幅度的修剪（LAMP）剪枝方案损失一定精度换取模型体积、运算量的降低，为下一步部署到嵌入式设备做好准备。实验结果表明，最终的改进模型相对于YOLOv7-tiny模型精度上提升3个百分点，FPS提升原来的119.4%，模型大小压缩到原来的14%。

关键词: 输电线路, YOLOv7-tiny算法, 通道剪枝, 异物检测

Abstract: In response to the problem of poor accuracy and the large model size in transmission line foreign object detection, an improved YOLOv7-tiny algorithm based on channel pruning has been proposed. Firstly, the ReXNet network is used to replace the backbone network of YOLOv7-tiny to address the feature bottleneck issue in the original network. Secondly, diversified branch blocks are introduced to enhance the network’s feature fusion capability. Finally, through layer-adaptive magnitude-based pruning (LAMP), a pruning approach is employed to trade off some accuracy for a reduction in model size and computational load, preparing it for deployment on embedded devices. Experimental results indicate that the final improved model achieves a 3?percentage points increase in accuracy compared to the YOLOv7-tiny model, a 119.4% increase in FPS, and compresses the model size to 14% of the original size.

Key words: transmission lines, YOLOv7-tiny algorithm, channel pruning, foreign object detection

孙阳, 李佳. 基于通道剪枝的YOLOv7-tiny输电线路异物检测算法[J]. 计算机工程与应用, 2024, 60(14): 319-328.

SUN Yang, LI Jia. YOLOv7-tiny Transmission Line Foreign Object Detection Algorithm Based on Channel Pruning[J]. Computer Engineering and Applications, 2024, 60(14): 319-328.

参考文献

[1] HUANG X B, LI H B, ZHU Y C. Short-term ice accretion fore casting model for transmission lines with modified time-series analysis by fireworks algorithm[J]. IET Generation, Transmission & Distribution, 2018, 12(5): 1074-1080.
[2] 陈杰, 朱仕焜, 孙嫱, 等. 面向无人机前端轻量级应用的输电线路鸟巢智能检测[J]. 福州大学学报 (自然科学版), 2023, 51(4): 539-546.
CHEN J, ZHU S K, SUN Q. Bird’s nest intelligent detection on transmission lines for unmanned aerial vehicles front-end lightweight application[J]. Journal of Fuzhou University(Natural Science Edition), 2023, 51(4) : 539-546.
[3] 王宇博, 尚军利, 张烨, 等. 基于改进YOLOv7的实时输电导线缺陷检测方法[J]. 南方电网技术, 2023, 17(12): 127-134.
WANG Y B, SHANG J L, ZHANG Y, et al. Real-time transmission wire defect detection method based on improved YOLOv7[J]. Southern Power System Technology, 2023, 17(12): 127-134.
[4] 杨锋, 丁之桐, 邢蒙蒙, 等. 深度学习的目标检测算法改进综述[J]. 计算机工程与应用, 2023, 59(11): 1-15.
YANG F, DING Z T, XING M M, et al. Review of object detection algorithm improvement in deep learning[J]. Computer Engineering and Applications, 2023, 59(11): 1-15.
[5] 邵瑰玮, 刘壮, 付晶, 等. 架空输电线路无人机巡检技术研究进展[J]. 高电压技术, 2020, 46(1): 14-22.
SHAO G W, LIU Z, FU J, et al. Research progress in unmanned aerial vehicle inspection technology on overhead transmission lines[J]. High Voltage Engineering, 2020, 46(1): 14-22.
[6] GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2014: 580-587.
[7] 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, 2016: 779-788.
[8] LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot multibox detector[C]//Proceedings of the European Conference on Computer Vision, 2016: 21-37.
[9] LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection[C]//Proceedings of the IEEE International Conference on Computer Vision, 2017: 2980-2988.
[10] WANG C Y, BOCHKOVSKIY A, LIAO H Y M. YOLOv7: trainable bag-of-freebies sets new state-of-the-art for realtime object detectors[J]. arXiv:2207.02696, 2022.
[11] 刘浩翰, 樊一鸣, 贺怀清, 等. 改进YOLOv7-tiny的目标检测轻量化模型[J]. 计算机工程与应用, 2023, 59(14): 166-175.
LIU H H, FAN Y M, HE H Q, et al. Improved YOLOv7-tiny’s object detection lightweight model[J]. Computer Engineering and Applications, 2023, 59(14): 166-175.
[12] 魏小玉, 焦良葆, 刘子恒, 等. 融合YOLOv3与改进ReXNet的手势识别方法研究[J]. 计算机测量与控制, 2023, 31(7): 278-289.
WEI X Y, JIAO L B, LIU Z H, et al. Research on gesture recognition method based on integrating YOLOv3 with improved ReXNet[J]. Computer Measurement & Control, 2023, 31(7): 278-289.
[13] DING X H, ZHANG X Y, HAN J G, et al. Diverse branch block: building a convolution as an inception-like unit[C]//Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021: 10881-10890.
[14] JAEHO L, SEJUN P, SANGWOO M, et al. Layer-adaptive sparsity for the magnitude-based pruning[J]. arXiv:2010. 07611, 2021.
[15] LI H, ASIM K, IGOR D, et al. Pruning filters for efficient convnets[J]. arXiv:1608.08710, 2016.
[16] HOWARD A, SANDLER M, CHEN B, et al. Searching for mobileNetV3[C]//Proceedings of the 2019 IEEE CVF International Conference on Computer Vision, 2020: 1314-1324.
[17] DAVIS B, JOSE J G O, JPNATHAN F, et al. What is the state of neural network pruning[J]. arXiv:2003.03033, 2020.
[18] PAVLO M, STEPHEN T, TERO K, et al. Pruning convolutional neural networks for resource efficient inference[J]. arXiv:1611.06440, 2016.
[19] HOWARD A G, ZHU M, CHEN B, et al. MobileNets: efficient convolutional neural networks for mobile vision applications[J]. arXiv:1704.04861, 2017.
[20] LIU Z. Learning efficient convolutional networks through network slimming[C]//Proceedings of the 2017 IEEE International Conference on Computer Vision, 2017: 2755-2763.
[21] HONGRONG C, MIAO Z, JAVEN Q S. A survey on deep neural network pruning-taxonomy, comparison, analysis, and recommendations[J]. arXiv:2308.06767, 2023.
[22] FANG G F, MA X Y, SONG M L, et al. DepGraph: towards any structural pruning[C]//Proceedings of the IEEE CVF Conference on Computer Vision and Pattern Recognition, 2023: 16091-1610.