基于改进YOLOv7的MODF端口状态检测算法

doi:10.19678/j.issn.1000-3428.0068445

摘要/Abstract

摘要：

人工巡检的管理方式导致光纤总配线架(MODF)端口状态的信息准确率较低, 无法区分占用端口与虚占端口。针对MODF资源管理中的端口状态识别问题, 提出一种改进的YOLOv7目标检测模型。鉴于数据集采集困难且类别不均衡, 采用多种数据增强方法来扩充数据集; 在骨干网络中使用共享权重的感受野扩大模块(RFEM), 扩大端口目标的感受野, 减少训练过程中的过拟合风险; 提出F-EMA注意力模块, 以提高对空间上下文信息的利用率, 减少因端口接近或被遮挡而导致的漏检、误检等情况; 使用NWD损失函数替代交并比(IoU)度量, 减轻对小目标位置偏差的敏感性, 提升密集小物体检测准确率。实验结果表明, 改进模型的mAP@0.5值达到98.8%, 相比原Yolov7模型提升了2百分点, mAP@0.5∶0.95值达到63.8%, 提升了9.5百分点, 提高了MODF端口资源利用率, 满足智能巡检系统对于端口占用状态识别准确率的基本要求。

关键词: 深度学习, YOLOv7算法, 光纤总配线架, 损失函数, 感受野扩大模块, 注意力模块

Abstract:

In recent years, manual inspection management methods have led to low accuracy in identifying the status of Fiber Optic Distribution Frame (MODF) ports, making it difficult to differentiate between occupied and unoccupied ports. To address the problem of status recognition in MODF port resource management, this study proposes an improved YOLOv7 object-detection model. First, owing to the difficulty in data collection and unbalanced categories, multiple data enhancement methods are used to expand the dataset. In addition, a shared-weight Receptive Field Expansion Module (RFEM) is used in the backbone network to enlarge the receptive field of the port targets and reduce the risk of overfitting during the training process. The F-EMA attention module is proposed to improve the utilization of spatial context information and reduce missed detections and false alarms caused by ports being closed or occluded. The Normalized Gaussian Wasserstein Distance (NWD) loss function is used to replace the Intersection over Union (IoU) measurement, which alleviates the sensitivity to the position deviation of small targets and improves the detection accuracy of dense small objects. The experimental results show that the mAP@0.5 value of the improved model reaches 98.8%, which is 2 percentage points higher than that of the original Yolov7 model, whereas the mAP@0.5∶0.95 value reaches 63.8%, which is 9.5 percentage points higher. This improves the utilization rate of MODF port resources and meets the basic requirements of the intelligent inspection system for the accuracy of port occupancy status recognition.

Key words: deep learning, YOLOv7 algorithm, Fiber Optic Distribution Frame (MODF), loss function, Receptive Field Expansion Module (RFEM), attention module

胡朝举, 郭凤仪. 基于改进YOLOv7的MODF端口状态检测算法[J]. 计算机工程, 2025, 51(2): 78-85.

HU Chaoju, GUO Fengyi. MODF Port State Detection Algorithm Based on Improved YOLOv7[J]. Computer Engineering, 2025, 51(2): 78-85.

https://www.ecice06.com/CN/Y2025/V51/I2/78

图/表 12

图1 Yolov7网络结构

Fig.1 Yolov7 network structure

图2 RFEM结构

Fig.2 RFEM structure

图3 EMA结构

Fig.3 EMA structure

图4 二维漏斗状条件T(X)

Fig.4 Two-dimensional funnel-shaped condition T(X)

图5 数据增强图片

Fig.5 Data enhanced images

图6 Loss曲线

Fig.6 Loss curve

图7 检测结果对比

Fig.7 Comparison of detection results

参考文献 25

1	朱明亮, 顾秀秀, 胡梓义, 等. 新运维时代MODF光纤配线架应用场景研究. 信息技术与信息化, 2020, (5): 117- 119.
	ZHU M L , GU X X , HU Z Y . Research on application scenarios of MODF optical distribution frame in new operation and maintenance era. Information Technology and Informatization, 2020, (5): 117- 119.
2	MENG Z Y , XU S Z , WANG L C , et al. Defect object detection algorithm for electroluminescence image defects of photovoltaic modules based on deep learning. Energy Science & Engineering, 2022, 10 (3): 800- 813. doi: 10.1002/ese3.1056
3	EL-SHAFAI W , ALI A M , EL-RABAIE E S M , et al. Automated COVID-19 detection based on single-image super-resolution and CNN models. Computers, Materials & Continua, 2022, 70 (1): 1141- 1157. doi: 10.32604/cmc.2022.018547
4	WANG H , XU Y S , WANG Z N , et al. CenterNet-auto: a multi-object visual detection algorithm for autonomous driving scenes based on improved CenterNet. IEEE Transactions on Emerging Topics in Computational Intelligence, 2023, 7 (3): 742- 752. doi: 10.1109/TETCI.2023.3235381
5	ZHANG L G , WANG L , JIN M , et al. Small object detection in remote sensing images based on attention mechanism and multi-scale feature fusion. International Journal of Remote Sensing, 2022, 43 (9): 3280- 3297. doi: 10.1080/01431161.2022.2089539
6	PLATT J C . A fast algorithm for training support vector machines. Journal of Information Technology, 1998, 2 (5): 1- 28.
7	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. Washington D. C., USA: IEEE Press, 2014: 580-587. 10.1109/CVPR.2014.81
8	GIRSHICK R. Fast R-CNN[C]//Proceedings of the IEEE International Conference on Computer Vision(ICCV). Washington D. C., USA: IEEE Press, 2015: 1440-1448.
9	REN S Q , HE K M , GIRSHICK R , et al. Faster R-CNN: towards real-time object detection with region proposal networks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39 (6): 1137- 1149. doi: 10.1109/TPAMI.2016.2577031
10	LI G Q , HAO X Y , ZHA L L , et al. An outstanding adaptive multi-feature fusion YOLOv3 algorithm for the small target detection in remote sensing images. Pattern Analysis and Applications, 2022, 25 (4): 951- 962. doi: 10.1007/s10044-022-01072-5
11	窦允冲, 侯进, 曾雷鸣, 等. 基于反馈机制与空洞卷积的道路小目标检测网络. 计算机工程, 2023, 49 (1): 287- 294. doi: 10.19678/j.issn.1000-3428.0063575
	DOU Y C , HOU J , ZENG L M , et al. Road small target detection network based on feedback mechanism and dilated convolution. Computer Engineering, 2023, 49 (1): 287- 294. doi: 10.19678/j.issn.1000-3428.0063575
12	REDMON J, FARHADI A. YOLO9000: better, faster, stronger[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Washington D. C., USA: IEEE Press, 2017: 6517-6525. 10.1109/CVPR.2017.690
13	戚玲珑, 高建瓴. 基于改进YOLOv7的小目标检测. 计算机工程, 2023, 49 (1): 41- 48. doi: 10.19678/j.issn.1000-3428.0065942
	QI L L , GAO J L . Small object detection based on improved YOLOv7. Computer Engineering, 2023, 49 (1): 41- 48. doi: 10.19678/j.issn.1000-3428.0065942
14	LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot MultiBox detector[C]//Proceedings of European Conference on Computer Vision. Berlin, Germany: Springer, 2016: 21-37. 10.1007/978-3-319-46448-0_2
15	苏东, 余宁梅. 基于深度学习的OBD端口占用状态自动识别算法. 北京邮电大学学报, 2019, 42 (6): 49- 57. doi: 10.13190/j.jbupt.2019-180
	SU D , YU N M . Research on automatic recognition algorithm of OBD port occupancy state based on deep learning. Journal of Beijing University of Posts and Telecommunications, 2019, 42 (6): 49- 57. doi: 10.13190/j.jbupt.2019-180
16	何景晖, 敖银辉, 赵伟良. 光缆交接箱端口状态的视觉检测方法. 微处理机, 2021, 42 (2): 53- 57. doi: 10.3969/j.issn.1002-2279.2021.02.014
	HE J H , AO Y H , ZHAO W L . A visual detection method for port state of optical cable cross connection cabinet. Microprocessors, 2021, 42 (2): 53- 57. doi: 10.3969/j.issn.1002-2279.2021.02.014
17	苏东. 无源光网络端口资源自动采集算法研究[D]. 西安: 西安理工大学, 2021. 10.27398/d.cnki.gxalu.2021.000024
	SU D. Research on automatic acquisition algorithm of port resources in passive optical network[D]. Xi'an: Xi'an University of Technology, 2021. (in Chinese)10.27398/d.cnki.gxalu.2021.000024
18	李蒙. 基于图像处理的光交箱状态识别和系统开发[D]. 广州: 广东工业大学, 2020. 10.27029/d.cnki.ggdgu.2020.002130
	LI M. State recognition and system development of optical box based on image processing[D]. Guangzhou: Guangdong University of Technology, 2020. (in Chinese)10.27029/d.cnki.ggdgu.2020.002130
19	黄奎. 基于计算机视觉的光纤传输箱端口状态识别与系统开发[D]. 广州: 广东工业大学, 2022. 10.27029/d.cnki.ggdgu.2022.001907
	HUANG K. Port state identification and system development of optical fiber transmission box based on computer vision[D]. Guangzhou: Guangdong University of Technology, 2022. (in Chinese)10.27029/d.cnki.ggdgu.2022.001907
20	WANG C Y, BOCHKOVSKIY A, LIAO H Y M. YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington D. C., USA: IEEE Press, 2023: 7464-7475. 10.1109/CVPR52729.2023.00721
21	YU Z P , HUANG H B , CHEN W J , et al. YOLO-FaceV2: a scale and occlusion aware face detector. Pattern Recognition, 2024, 155, 110714. doi: 10.1016/j.patcog.2024.110714
22	YU F, KOLTUN V. Multi-scale context aggregation by dilated convolutions[EB/OL]. (2016-03-04)[2023-09-21]. https://doi.org/10.48550/arXiv.1511.07122.
23	OUYANG D L, HE S, ZHANG G Z, et al. Efficient multi-scale attention module with cross-spatial learning[C]//Proceedings of 2023 IEEE International Conference on Acoustics, Speech and Signal Processing(ICASSP). Washington D. C., USA: IEEE Press, 2023: 1-5. 10.1109/ICASSP49357.2023.10096516
24	QIU S, XU X M, CAI B L. FReLU: flexible rectified linear units for improving convolutional neural networks[C]//Proceedings of the 24th International Conference on Pattern Recognition (ICPR). Washington D. C., USA: IEEE Press, 2018: 1223-1228. 10.1109/ICPR.2018.8546022
25	WANG J W, XU C, YANG W, et al. A Normalized Gaussian Wasserstein Distance for Tiny Object Detection[EB/OL]. (2022-06-14)[2023-09-21]. https://doi.org/10.48550/arXiv.2110.13389.

[1]	孙浩淼, 李宗民, 肖倩, 孙文洁, 张雯欣. AI-Curling: 一种冰壶现场分析与决策方法[J]. 计算机工程, 2025, 51(2): 102-110.
[2]	杨旺达, 万亚平, 邹刚, 闵晓珊, 王沂, 陆宇程. 驾驶素质缺失测试眼状态的深度学习分类方法研究[J]. 计算机工程, 2025, 51(2): 149-158.
[3]	孙义康, 高建华. 基于卷积神经网络和长短期记忆的死代码检测方法[J]. 计算机工程, 2025, 51(2): 223-237.
[4]	龙丽叶, 焦世超, 郭磊, 韩燮, 况立群. 基于紧凑中心的多模态三维模型检索研究[J]. 计算机工程, 2025, 51(2): 322-334.
[5]	周宇, 谢威, 邝得互, 江健民. 基于三元自注意力的视频快照压缩成像重建[J]. 计算机工程, 2025, 51(1): 20-30.
[6]	胡升龙, 陈彬, 张开华, 宋慧慧. 场景结构知识增强的协同显著性目标检测[J]. 计算机工程, 2025, 51(1): 31-41.
[7]	喻勇涛, 孙奥, 李昂, 朱琳琳. 基于孪生网络的分类器输出重复性优化方法[J]. 计算机工程, 2025, 51(1): 118-127.
[8]	张会影, 圣文顺. 基于标记适应的人脸年龄识别优化算法[J]. 计算机工程, 2025, 51(1): 174-181.
[9]	杨红菊, 吉昌. 学习驱动的图像压缩算法研究[J]. 计算机工程, 2025, 51(1): 190-197.
[10]	赵南南, 高翡晨. 基于改进YOLOv8的交通场景实例分割算法[J]. 计算机工程, 2025, 51(1): 198-207.
[11]	王晓路, 汶建荣. 基于运动-时间感知的人体动作识别方法[J]. 计算机工程, 2025, 51(1): 216-224.
[12]	火久元, 苏泓瑞, 武泽宇, 王婷娟. 基于改进YOLOv8的道路交通小目标车辆检测算法[J]. 计算机工程, 2025, 51(1): 246-257.
[13]	王骞, 张俊华, 王泽彤, 李博. X2S-Net: 基于双平面X线片的脊柱三维重建[J]. 计算机工程, 2025, 51(1): 277-286.
[14]	易鹏, 杨晔, 严仕嘉. 基于MPCNN模型的sEMG快速迁移学习的手势识别应用研究[J]. 计算机工程, 2025, 51(1): 304-311.
[15]	刘兆伟, 方艳红, 郑明宇, 锁斌. 基于注意力机制与多任务的肺部疾病诊断方法[J]. 计算机工程, 2025, 51(1): 332-342.

选择文件类型/文献管理软件名称

选择包含的内容