基于改进元学习的小样本钢丝绳表面损伤检测

doi:10.3778/j.issn.1002-8331.2401-0359

摘要/Abstract

摘要： 为了解决构建包含不同形式钢丝绳表面损伤的数据集的困难以及数据量的一致性所带来的挑战，引入了一种基于改进元学习的小样本钢丝绳表面损伤检测的新方法。在该方法中，在查询分支中使用可变形卷积代替传统卷积，从而便于提取更细微的空间信息。此外，在颈部网络中加入双向特征金字塔网络来合并查询特征。在支持分支的设计中，构建了全局精细化网络，从支持图像中提取类别特征向量，并采用自适应条件池化机制对特征图像进行压缩。为了有效地提取查询样本中的感兴趣区域，引入了一个多尺度重加权模块，利用改进的ECA通道注意力-坐标注意力方法，并且将经过坐标注意力得到的输出送入到Re-SPPD模块。在此基础上，设计了专用的小目标检测层，提高了识别小目标损伤情况的精度。实验结果表明，与其他方法相比，该方法在钢丝绳数据集上的5-shot，10-shot任务的平均精度（mAP）分别为76.15%和83.25%。这些结果明显优于其他方法，肯定了所提出的方法的有效性。

关键词: 元学习, 小样本学习, 变形卷积, 坐标注意力, 钢丝绳, 小目标检测层

Abstract: Addressing the challenges associated with the difficulty of constructing a dataset encompassing diverse forms of surface damage on wire ropes and the uniformity in data volume, this paper introduces a novel approach for few-shot surface damage detection in wire ropes, leveraging an enhanced meta-learning framework. Within this methodology, deformable convolution is employed in lieu of conventional convolution within the query branch, facilitating the extraction of more nuanced spatial information. Additionally, it incorporates a bidirectional feature pyramid network in the neck network to amalgamate the query features. In the design of the support branch, a global refinement network is formulated to derive category feature vectors from the supporting images, and a conditional pooling mechanism is implemented to compress the feature images. To effectively extract regions of interest in query samples, a multi-scale weighted module is introduced, utilizing an improved channel attention-coordinate attention approach, and the output obtained through the coordinate attention is fed into the Re-SPPD module. Furthermore, a dedicated small target detection layer is devised to discern the damage condition of smaller entities with heightened precision. Experimental findings demonstrate that, in comparison to alternative methodologies, the proposed approach attains an average mean average precision (mAP) of 76.15% and 83.25% for 5-shot and 10-shot tasks, respectively, on the wire rope dataset. These results significantly outperform other methods, affirming the efficacy of the proposed methodology.

Key words: meta-learning, few-shot learning, deformation convolution, coordinate attention, wire rope, small target detection layer

黄毅, 郭金梦, 任子誉, 张金来, 任广安, 付玲. 基于改进元学习的小样本钢丝绳表面损伤检测[J]. 计算机工程与应用, 2025, 61(13): 256-269.

HUANG Yi, GUO Jinmeng, REN Ziyu, ZHANG Jinlai, REN Guang’an, FU Ling. Few-Shot Surface Damage Detection of Wire Rope Using Improved Meta-Learning[J]. Computer Engineering and Applications, 2025, 61(13): 256-269.

参考文献

[1] LIU S W, SUN Y H, JIANG X Y, et al. Comparison and analysis of multiple signal processing methods in steel wire rope defect detection by hall sensor[J]. Measurement, 2021, 171: 108768.
[2] KIKUCHI H, TSCHUNCKY R, SZIELASKO K. Challenges for detection of small defects of submillimeter size in steel using magnetic flux leakage method with higher sensitive magnetic field sensors[J]. Sensors and Actuators A: Physical, 2019, 300: 111642.
[3] EFTEKHARI H, TEHRANCHI M M. Miniaturized magneto-optical imaging sensor for crack and micro-crack detection[J]. Optik, 2020, 207: 163830.
[4] TOGO R, WATANABE H, OGAWA T, et al. Deep convolutional neural network-based anomaly detection for organ classification in gastric X-ray examination[J]. Computers in Biology and Medicine, 2020, 123: 103903.
[5] XU Y T, WU Z F, ZHANG H X, et al. Detection of prohibited and restricted object X-ray images based on Hi color space segmentation[J]. Journal of X-Ray Science and Technology, 2023, 31(5): 1093-1114.
[6] BENSAID N, BENLAMNOUAR M F, BADJI R, et al. Microstructural and mechanical characterization of HSLA X70 welded joints using eddy current testing (ECT)[J]. IOP Conference Series: Materials Science and Engineering, 2022, 1233(1): 012012.
[7] KYRKOU C. YOLOpeds: efficient real-time single-shot pedestrian detection for smart camera applications[J]. IET Computer Vision, 2020, 14(7): 417-425.
[8] CHIAN E, FANG W L, GOH Y M, et al. Computer vision approaches for detecting missing barricades[J]. Automation in Construction, 2021, 131: 103862.
[9] EL HAKEA A H, FAKHR M W. Recent computer vision applications for pavement distress and condition assessment[J]. Automation in Construction, 2023, 146: 104664.
[10] CAO Y N, ZHANG D L, WANG C, et al. A novel electromagnetic method for local defects inspection of wire rope[C]//Proceedings of the TENCON 2006-2006 IEEE Region 10 Conference. Piscataway: IEEE, 2006: 1-4.
[11] HUANG X Y, LIU Z L, ZHANG X Y, et al. Surface damage detection for steel wire ropes using deep learning and computer vision techniques[J]. Measurement, 2020, 161: 107843.
[12] YE H L, WANG Y, CAO F L. A novel meta-learning framework: multi-features adaptive aggregation method with information enhancer[J]. Neural Networks, 2021, 144: 755-765.
[13] ZHANG Y, ZUO X, ZHENG X X, et al. Improving metric-based few-shot learning with dynamically scaled softmax loss[J]. Image and Vision Computing, 2023, 140: 104860.
[14] WANG H, TIAN S Z, FU Y, et al. Feature augmentation based on information fusion rectification for few-shot image classification[J]. Scientific Reports, 2023, 13(1): 3607.
[15] SCHWARTZ E, LEONID K, JOSEPH S, et al. RepMet: representative-based metric learning for classification and one-shot object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018.
[16] XIAO Z X, QI J H, XUE W, et al. Few-shot object detection with self-adaptive attention network for remote sensing images[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2021, 14: 4854-4865.
[17] CHEN H, WANG Y L, WANG G Y, et al. LSTD: a low-shot transfer detector for object detection[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2018.
[18] HU H Z, BAI S, LI A X, et al. Dense relation distillation with context-aware aggregation for few-shot object detection[C]//Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2021: 10180-10189.
[19] GAO Y, LI H J, FU W Q. Few-shot learning for image-based bridge damage detection[J]. Engineering Applications of Artificial Intelligence, 2023, 126: 107078.
[20] WANG B, JIANG P, GAO J X, et al. A lightweight few-shot marine object detection network for unmanned surface vehicles[J]. Ocean Engineering, 2023, 277: 114329.
[21] FINN C, ABBEEL P, LEVINE S. Model-agnostic meta-learning for fast adaptation of deep networks[J]. arXiv:1703.03400, 2017.
[22] WANG C Y, BOCHKOVSKIY A, LIAO H M. YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]//Proceedings of the 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2023: 7464-7475.
[23] DAI J F, QI H Z, XIONG Y W, et al. Deformable convolutional networks[C]//Proceedings of the 2017 IEEE International Conference on Computer Vision. Piscataway: IEEE, 2017: 764-773.
[24] ZHANG J L, MENG Y M, WU J F, et al. Monitoring sugar crystallization with deep neural networks[J]. Journal of Food Engineering, 2020, 280: 109965.
[25] WEI H J, XU E Y, ZHANG J L, et al. BushNet: effective semantic segmentation of bush in large-scale point clouds[J]. Computers and Electronics in Agriculture, 2022, 193: 106653.
[26] YANG W J, WU J C, ZHANG J L, et al. Deformable convolution and coordinate attention for fast cattle detection[J]. Computers and Electronics in Agriculture, 2023, 211: 108006.
[27] CHILUKURI D M, YI S, SEONG Y. A robust object detection system with occlusion handling for mobile devices[J]. Computational Intelligence, 2022, 38(4): 1338-1364.
[28] TAN M X, PANG R M, LE Q V. EfficientDet: scalable and efficient object detection[C]//Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 10778-10787.
[29] BAYRAKTAR E, YIGIT C B. Conditional-pooling for improved data transmission[J]. Pattern Recognition, 2024, 145: 109978.
[30] CHEN W, HE Z L, ZHANG J. Online monitoring of crack dynamic development using attention-based deep networks[J]. Automation in Construction, 2023, 154: 105022.
[31] WANG Q L, WU B G, ZHU P F, 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: 11531-11539.
[32] HOU Q B, ZHOU D Q, FENG J S. Coordinate attention for efficient mobile network design[C]//Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2021: 13708-13717.
[33] 孟莉莎, 杨贤昭, 刘惠康. 基于CA-EfficientNetV2的蘑菇图像分类算法研究[J]. 激光与光电子学进展, 2022, 59(24): 56-63.
MENG L S, YANG X Z, LIU H K. Algorithm on mushroom image classification based on CA-EfficientNetV2[J]. Laser & Optoelectronics Progress, 202, 59(24): 56-63.
[34] GUO W, QIAO S, ZHAO C Y, et al. Defect detection for industrial neutron radiographic images based on modified YOLO network[J]. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 2023, 1056: 168694.
[35] CHENG G, YUAN X, YAO X W, et al. Towards large-scale small object detection: survey and benchmarks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023, 45(11): 13467-13488.
[36] CHEN W Y, LIU Y C, Kira Z, et al. A closer look at few-shot classification[J]. arXiv:1904.04232, 2019.
[37] REZATOFIGHI H, TSOI N, GWAK J, et al. Generalized intersection over union: a metric and a loss for bounding box regression[C]//Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2019: 658-666.
[38] KANG B Y, LIU Z, WANG X, et al. Few-shot object detection via feature reweighting[C]//Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2019: 8419-8428.
[39] LI S F, LI K Y, QIAO Y, et al. A multi-scale cucumber disease detection method in natural scenes based on YOLOv5[J]. Computers and Electronics in Agriculture, 2022, 202: 107363.
[40] LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot multibox detector[C]//Proceedings of the 14th European Conference on Computer Vision. Cham: Springer, 2016: 21-37.
[41] 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.
[42] LI Y W, FENG W Q, LYU S C, et al. Feature reconstruction and metric based network for few-shot object detection[J]. Computer Vision and Image Understanding, 2023, 227: 103600.
[43] TSENG F H, YEH K H, KAO F Y, et al. MiniNet: dense squeeze with depthwise separable convolutions for image classification in resource-constrained autonomous systems[J]. ISA Transactions, 2023, 132: 120-130.
[44] DONG Y P, KANG C X, ZHANG J L, et al. Benchmarking robustness of 3D object detection to common corruptions in autonomous driving[C]//Proceedings of the 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2023: 1022-1032.
[45] CHOLLET F. Xception: deep learning with depthwise separable convolutions[C]//Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2017: 1800-1807.
[46] PANG J M, CHEN K, SHI J P, et al. Libra R-CNN: towards balanced learning for object detection[C]//Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2019: 821-830.
[47] HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]//Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2018: 7132-7141.
[48] WOO S, PARK J, LEE J Y, et al. CBAM: convolutional block attention module[C]//Proceedings of the European Conference on Computer Vision. Cham: Springer, 2018: 3-19.
[49] SELVARAJU R R, DAS A, VEDANTAM R, et al. Grad-CAM: why did you say that? Visual explanations from deep networks via gradient-based localization[J]. arXiv:1610. 02391, 2016.