改进YOLOX算法的碳滑板缺陷检测方法

doi:10.3778/j.issn.1002-8331.2308-0023

摘要/Abstract

摘要： 针对碳滑板缺陷图像中存在大量密集小目标缺陷，容易出现误检及漏检的问题，基于YOLOX算法提出一种改进的YOLOX碳滑板缺陷检测算法。在主干网络中使用膨胀卷积，增大特征图的感受野，解决主干网络特征提取能力不足的问题；为了进一步降低损失，在YOLOX中构建新的连接并加入CBAM注意力模块，减少小目标缺陷漏检，检测能力也进一步提升；在颈部引入自适应特征融合网络ASFF，将不同尺度特征层进行融合，增强网络特征表达能力；用EIOU损失函数代替IOU损失函数，充分考虑了预测框和真实框纵横比对模型优化的影响，加快模型收敛速度的同时精度也有所提升。在自制碳滑板数据集上的实验结果表明，相比于原始YOLOX算法平均精确率提升4.89个百分点，对小目标识别率明显提升。

关键词: 缺陷检测, YOLOX, 注意力机制, 损失函数, 小目标检测

Abstract: In view of the large number of dense small target defects in the carbon slide defect image, it is easy to misdetect and miss the problem, an improved YOLOX carbon slide defect detection algorithm is proposed based on YOLOX algorithm. Firstly, expansive convolution is used in the backbone network to increase the receptive field of the feature map to solve the problem of insufficient feature extraction capability in the backbone network. Secondly, in order to further reduce the loss, new connections are built in YOLOX and CBAM attention module is added to reduce the missed detection of small target defects, and the detection capability is further improved. Thirdly, the adaptive feature fusion network ASFF is introduced in the neck to fuse the feature layers of different scales to enhance the feature expression ability of the network. Finally, the EIOU loss function is used to replace the IOU loss function, and the influence of the aspect ratio between the predicted frame and the real frame is fully considered. The convergence speed of the model is accelerated and the accuracy is improved. The experimental results on the self-made carbon slide data set show that compared with the original YOLOX algorithm, the average accuracy is improved by 4.89?percentage points, and the recognition rate of small targets is significantly improved.

Key words: defect detection, YOLOX, attention mechanism, loss function, small target detection

郑爱云, 蒋新宇, 刘伟民, 陈澍军, 郑直. 改进YOLOX算法的碳滑板缺陷检测方法[J]. 计算机工程与应用, 2024, 60(21): 244-253.

ZHENG Aiyun, JIANG Xinyu, LIU Weimin, CHEN Shujun, ZHENG Zhi. Improved YOLOX Algorithm for Carbon Contact Strip Detection Method[J]. Computer Engineering and Applications, 2024, 60(21): 244-253.

参考文献

[1] 周宁, 杨文杰, 刘久锐, 等. 基于受电弓状态感知的弓网安全监测系统研究与探讨[J]. 中国科学: 技术科学, 2021, 51(1): 23-34.
ZHOU N, YANG W J, LIU J R, et al. Investigation of a pantograph-catenary monitoring system using condition-based pantograph recognition[J]. Scientia Sinica (Technologica), 2021, 51(1): 23-34.
[2] 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.
[3] REN S, HE K, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. arXiv:1506.01497, 2015.
[4] HE K M, GKIOXARI G, DOLLAR P. Mask R-CNN[C]//Proceedings of the 2017 IEEE International Conference on Computer Vision, Venice, Italy, 2017: 2980-2988.
[5] REDMON J, FARHADI A. YOLOv3: an incremental improvement[J]. arXiv:1804.02767, 2018.
[6] BOCHKOVSKIY A, WANG C Y, LIAO H Y M. YOLOv4: optimal speed and accuracy of object detection[J]. arXiv: 2004.10934, 2020.
[7] LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot multibox detector[C]//Proceedings of the 14th European Conference on Computer Vision (CCV 2016), Amsterdam, The Netherlands, October 11-14, 2016. Cham: Springer International Publishing, 2016: 21-37.
[8] GE Z, LIU S, WANG F, et al. YOLOx: exceeding YOLO series in 2021[J]. arXiv:2107.08430, 2021.
[9] LI D, ZHANG Z, WANG B, et al. Detection method of timber defects based on target detection algorithm[J]. Measurement, 2022, 203: 111937.
[10] 庹冰, 黄丽雯, 唐鑫, 等. 基于YOLOX-WSC的PCB缺陷检测算法研究[J]. 计算机工程与应用, 2023, 59(10): 236-243.
TUO B, HUANG L W, TANG X, et al. Research on PCB defect detection algorithm based on YOLOX-WSC[J]. Computer Engineering and Applications, 2023, 59(10): 236-243.
[11] WANG X, GAO J S, HOU B J, et al. A lightweight modified YOLOX network using coordinate attention mechanism for PCB surface defect detection[J]. IEEE Sensors Journal, 2022, 22(21): 20910-20920.
[12] 李亚东, 马行, 穆春阳, 等. 改进YOLOX网络的轴承缺陷小目标检测方法[J]. 计算机工程与应用, 2023, 59(1): 100-107.
LI Y D, MA X, MU C Y, et al. Improved small target detection method of bearing defects in YOLOX network[J]. Computer Engineering and Applications, 2023, 59(1): 100-107.
[13] GUO Q, LIU J, KALIUZHNYI M. YOLOX-SAR: high-precision object detection system based on visible and infrared sensors for SAR remote sensing[J]. IEEE Sensors Journal, 2022, 22(17): 17243-17253.
[14] YI C, XU B, CHEN J, et al. An improved YOLOX model for detecting strip surface defects[J]. Steel Research International, 2022, 93(11): 2200505.
[15] HU J, QIAO P, LV H, et al. High speed railway fastener defect detection by using improved YOLOX-Nano model[J]. Sensors, 2022, 22(21): 8399.
[16] WEI X, JIANG S, LI Y, et al. Defect detection of pantograph slide based on deep learning and image processing technology[J]. IEEE Transactions on Intelligent Transportation Systems, 2019, 21(3): 947-958.
[17] LU S, LIU Z, LI D, et al. Automatic wear measurement of pantograph slider based on multiview analysis[J]. IEEE Transactions on Industrial Informatics, 2020, 17(5): 3111-3121.
[18] LI H. Research on fault detection algorithm of pantograph based on edge computing image processing[J]. IEEE Access, 2020, 8: 84652-84659.
[19] KARADUMAN G, AKIN E. A deep learning based method for detecting of wear on the current collector Strips’ surfaces of the pantograph in railways[J]. IEEE Access, 2020, 8: 183799-183812.
[20] CHEN L C, PAPANDREOU G, SCHROFF F, et al. Rethinking atrous convolution for semantic image segmentation[J]. arXiv:1706.05587, 2017.
[21] LIU S, QI L, QIN H, et al. Path aggregation network for instance segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018: 8759-8768.
[22] WOO S, PARK J, LEE J Y, et al. CBAM: convolutional block attention module[C]//Proceedings of the European Conference on Computer Vision (ECCV), 2018: 3-19.
[23] LIU S, HUANG D, WANG Y. Learning spatial fusion for single-shot object detection[J]. arXiv:1911.09516, 2019.
[24] JIANG B, LUO R, MAO J, et al. Acquisition of localization confidence for accurate object detection[C]//Proceedings of the European Conference on Computer Vision (ECCV), 2018: 784-799.
[25] REZATOFIGHI H, TSOI N, GWAK J Y, et al. Generalized intersection over union: a metric and a loss for bounding box regression[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019: 658-666.
[26] ZHENG Z, WANG P, LIU W, et al. Distance-IoU loss: faster and better learning for bounding box regression[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2020: 12993-13000.
[27] ZHANG Y F, REN W, ZHANG Z, et al. Focal and efficient IOU loss for accurate bounding box regression[J]. arXiv: 2101.08158, 2021.
[28] 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 lEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023: 7464-7475.
[29] ZHOU X, WANG D, KR?HENBüHL P. Objects as points[J]. arXiv:1904.07850, 2019.
[30] CARION N, MASSA F, SYNNAEVE G, et al. End-to-end object detection with transformers[C]//Proceedings of the European Conference on Computer Vision. Cham: Springer International Publishing, 2020: 213-229.