焊接是连接、支撑和强化各种结构和设备的重要技术，在承压管道设施中广泛应用。焊缝中若存在缺陷，会影响设备的使用性能甚至引发严重事故，因此，开展焊缝缺陷检测研究对于保障工业生产安全与效率具有重要意义。传统的X射线胶片中焊缝缺陷的判别方式为技术人员检视，效率低且检测结果易受人为因素影响。近年来，基于深度学习的X射线图像焊缝缺陷检测技术因其具有精测精度较高、检测速度快且无需手动设计特征提取网络的优势成为研究热点。本论文围绕实际工业中，焊缝缺陷图像公开数据集数量较少、类别不平衡的问题，以及焊缝缺陷检测算法精度低的问题，开展了基于深度学习的X射线图像焊缝缺陷检测方法研究。论文的主要研究工作如下：

针对公开数据集数量较少、类别不平衡的问题，本论文采集、处理并标注了1709张承压管道焊缝缺陷图像，形成了初始数据集WeldDefect，其中缺陷标注实例共2813个。为扩充WeldDefect数据集中稀缺的裂纹、夹渣和未熔合缺陷，展开了基于深度学习的图像生成算法研究，提出了WD-DCGAN网络，将总样本数量增加到2280张，其中缺陷标注实例为4047个，形成规模更大、类别更均衡的数据集WeldDefect-K。实验结果表明，数据集扩充后，裂纹、夹渣和未熔合缺陷的检测精度分别提升了10%、2.5%和3.2%，数据集总体识别精度提升了1.4%，为焊缝缺陷检测算法的研究提供了高质量数据集支持。

针对现有焊缝缺陷检测算法精度低的问题，本论文展开了基于深度学习的目标检测算法研究，提出了一种基于Transformer的HTS-DETR焊缝缺陷检测算法。以RT-DETR为基线模型，通过引入特征筛选融合模块、跨维度交互三重注意力机制以及轻量化GSConv和VoV-GSCSP模块，改进了网络的颈部结构，增强其对小尺寸目标的表征能力，提高模型精度的同时降低了计算复杂度。实验结果表明，HTS-DETR模型在PASCAL VOC数据集上，AP值达到68.4%，AP_s值达29.2%，均优于其他主流目标检测算法；在WeldDefect-K数据集上的mAP50值达到了86.3%，比基线网络提高了4.5%，同时参数量降低了18%，更适用于实际工业应用。

针对人工评片效率低的问题，本文结合HTS-DETR模型，开发了一套X射线图像焊缝缺陷自动检测系统，具有注册与登录、图像上传查看、图像预处理、图像检测及缺陷信息管理五大功能。实验结果表明，该系统检测速度达到136张/秒，平均检测精度达到了86.3%，能够显著提高焊缝缺陷的检测效率。

Welding is an essential technology for connecting, supporting, and strengthening various structures and equipment, and it is widely used in pressurized pipeline facilities. If defects exist in the weld seam, it can affect the performance of the equipment and even lead to serious accidents. Therefore, research on weld defect detection is of great significance for ensuring industrial production safety and efficiency. In traditional X-ray film, weld defect detection is carried out through manual inspection by technicians, which is inefficient and the results are easily influenced by human factors. In recent years, deep learning-based X-ray image weld defect detection technology has become a research hotspot due to its high precision, fast detection speed, and the advantage of not requiring manual design of feature extraction networks. This thesis focuses on the issues of limited public weld defect image datasets and class imbalance, as well as the low accuracy of weld defect detection algorithms in practical industrial applications. The research on deep learning-based X-ray image weld defect detection methods is conducted. The main work in this thesis is as follows:

To address the problem of limited public dataset quantity and class imbalance, this thesis collected, processed, and labeled 1,709 images of weld defects in pressurized pipelines, forming the initial dataset WeldDefect, which contains 2,813 labeled defect instances. To expand the underrepresented defects such as cracks, slag inclusion, and lack of fusion in the WeldDefect dataset, a deep learning-based image generation algorithm was developed, and the WD-DCGAN network was proposed. This increased the total number of samples to 2,280, with 4,047 defect instances, forming a larger and more balanced dataset, WeldDefect-K. Experimental results show that after dataset expansion, the detection accuracy of cracks, slag inclusion, and lack of fusion defects increased by 10%, 2.5%, and 3.2%, respectively, and the overall recognition accuracy of the dataset increased by 1.4%, providing high-quality dataset support for weld defect detection algorithm research.

To address the problem of low accuracy in existing weld defect detection algorithms, this thesis conducted research on deep learning-based object detection algorithms and proposed a Transformer-based HTS-DETR weld defect detection algorithm. Using RT-DETR as the baseline model, the network's neck structure was improved by introducing a feature selection and fusion module, a cross-dimensional interaction triple attention mechanism, and lightweight GSConv and VoV-GSCSP modules. These improvements enhanced the network's ability to represent small-sized objects, improved model accuracy, and reduced computational complexity. Experimental results show that the HTS-DETR model achieved an AP of 68.4% and an APs of 29.2% on the PASCAL VOC dataset, outperforming other mainstream object detection algorithms. On the WeldDefect-K dataset, the mAP50 value reached 86.3%, a 4.5% improvement over the baseline network, while the parameter count was reduced by 18%, making it more suitable for practical industrial applications.

To address the issue of low manual inspection efficiency, this thesis, in combination with the HTS-DETR model, developed an automatic X-ray image weld defect detection system. The system includes five key functions: registration and login, image upload and viewing, image preprocessing, image detection, and defect information management. Experimental results show that the system achieves a detection speed of 136 images per second, with an average detection accuracy of 86.3%, significantly improving the efficiency of weld defect detection.

[1]唐志成. 无损检测技术在承压类特种设备检验检测中的应用[J]. 中国质量监管, 2024, (06): 76-77.

[2]李少波, 杨静, 王铮, 等. 缺陷检测技术的发展与应用研究综述[J]. 自动化学报, 2020, 46(11): 2319-2336.

[3]赵家炜, 蔺健宁. 特种设备检验中无损检测技术的应用分析[J]. 中国设备工程, 2024, (21): 153-155.

[4]沈锦军, 罗展慧. 无损检测技术在压力容器和压力管道检验中的应用[J]. 设备监理, 2024, (03): 58-61.

[5]王睿, 高少泽, 刘卫朋, 等. 一种轻量级高效 X 射线焊缝图像缺陷检测方法[J]. 焊接学报, 2024, 45(7): 41-49.

[6]肖文凯, 南水鱼, 张琳琳. 基于卷积神经网络的X射线焊缝缺陷检测算法研究[J]. 自动化仪表, 2022, 43(08): 67-72.

[7]吴昉, 王伟, 刘卫朋. 结合注意力机制和卷积神经网络的 X 射线焊缝缺陷检测[J]. 科学技术与工程, 2023, 23(8): 3387-3395.

[8]王靖然, 王桂棠, 杨波, 等. 深度学习在焊缝缺陷检测的应用研究综述[J]. 机电工程技术, 2021, 50(3): 65-68.

[9]李超, 孙俊. 基于机器视觉方法的焊缝缺陷检测及分类算法[J]. 计算机工程与应用, 2018, 54(06): 264-270.

[10]李柯泉, 陈燕, 刘佳晨, 等. 基于深度学习的目标检测算法综述[J]. 计算机工程, 2022, 48(07): 1-12.

[11]张阳婷, 黄德启, 王东伟, 等. 基于深度学习的目标检测算法研究与应用综述[J]. 计算机工程与应用, 2023, 59(18): 1-13.

[12]李亚森, 李晔, 李赵辉. 基于深度学习的焊缝缺陷检测方法综述[J]. 焊接技术, 2024, 53(04): 6-13.

[13]Bharati P, Pramanik A. Deep learning techniques—R-CNN to mask R-CNN: a survey[J]. Computational Intelligence in Pattern Recognition: Proceedings of CIPR 2019, 2020: 657-668.

[14]Ren Z, Fang F, Yan N, et al. State of the art in defect detection based on machine vision[J]. International Journal of Precision Engineering and Manufacturing-Green Technology, 2022, 9(2): 661-691.

[15]Bi X, Hu J, Xiao B, et al. Iemask r-cnn: Information-enhanced mask r-cnn[J]. IEEE Transactions on Big Data, 2022, 9(2): 688-700.

[16]Li W. Analysis of object detection performance based on Faster R-CNN[C]. Journal of Physics: Conference Series. IOP Publishing, 2021, 1827(1): 012085.

[17]Tong K, Wu Y. Rethinking PASCAL-VOC and MS-COCO dataset for small object detection[J]. Journal of Visual Communication and Image Representation, 2023, 93(1): 103830.

[18]Zhang Y, Chi M. Mask-R-FCN: A deep fusion network for semantic segmentation[J]. IEEE Access, 2020, 8(8): 155753-155765.

[19]Wang S, Sun G, Zheng B, et al. A crop image segmentation and extraction algorithm based on mask RCNN[J]. Entropy, 2021, 23(9): 1160.

[20]Wu Y, Liu W, Wan S. Multiple attention encoded cascade R-CNN for scene text detection[J]. Journal of Visual Communication and Image Representation, 2021, 80: 103261.

[21]Jiang P, Ergu D, Liu F, et al. A Review of Yolo algorithm developments[J]. Procedia computer science, 2022, 199: 1066-1073.

[22]Terven J, Córdova-Esparza D M, Romero-González J A. A comprehensive review of yolo architectures in computer vision: From yolov1 to yolov8 and yolo-nas[J]. Machine Learning and Knowledge Extraction, 2023, 5(4): 1680-1716.

[23]Wang K, Liu M. YOLOv3-MT: A YOLOv3 using multi-target tracking for vehicle visual detection[J]. Applied Intelligence, 2022, 52(2): 2070-2091.

[24]Dewi C, Chen R C, Jiang X, et al. Deep convolutional neural network for enhancing traffic sign recognition developed on Yolo V4[J]. Multimedia Tools and Applications, 2022, 81(26): 37821-37845.

[25]He K, Zhang X, Ren S, et al. Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition[J].IEEE Transactions on Pattern Analysis & Machine Intelligence, 2014, 37(9):1904-16.

[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, 34(07): 12993-13000.

[27]Marmolin H. Subjective MSE measures[J]. IEEE transactions on systems, man, and cybernetics, 1986, 16(3): 486-489.

[28]Zhang Y, Guo Z, Wu J, et al. Real-time vehicle detection based on improved yolov5[J]. Sustainability, 2022, 14(19): 12274.

[29]Li Y, Zhang X. Object detection for uav images based on improved yolov6[J]. IAENG International Journal of Computer Science, 2023, 50(2): 759-768.

[30]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. 2023: 7464-7475.

[31]Sohan M, Sai Ram T, Reddy R, et al. A review on yolov8 and its advancements[C]. International Conference on Data Intelligence and Cognitive Informatics. Springer, Singapore, 2024: 529-545.

[32]Bakirci M, Bayraktar I. YOLOv9-Enabled Vehicle Detection for Urban Security and Forensics Applications[C]. 2024 12th International Symposium on Digital Forensics and Security (ISDFS). IEEE, 2024: 1-6.

[33]Cengil E. Weld Defect Detection with YOLOv10[J]. NATURENGS, 2024, 5(2): 77-81.

[34]Bakirci M, Dmytrovych P, Bayraktar I, et al. Multi-Class Vehicle Detection and Classification with YOLO11 on UAV-Captured Aerial Imagery[C]. 2024 IEEE 7th International Conference on Actual Problems of Unmanned Aerial Vehicles Development (APUAVD). IEEE, 2024: 191-196.

[35]Khan S, Naseer M, Hayat M, et al. Transformers in vision: A survey[J]. ACM computing surveys (CSUR), 2022, 54(10s): 1-41.

[36]Meng L, Li H, Chen B C, et al. Adavit: Adaptive vision transformers for efficient image recognition[C]. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2022: 12309-12318.

[37]Carion N, Massa F, Synnaeve G, et al. End-to-end object detection with transformers[C]. European conference on computer vision. Cham: Springer International Publishing, 2020: 213-229.

[38]Yin H, Chen L. Enhanced Road Vehicle Object Detection Based on Improved Deformable DETR[C]. 2024 5th International Seminar on Artificial Intelligence, Networking and Information Technology (AINIT). IEEE, 2024: 2227-2230.

[39]Sun Z, Cao S, Yang Y, et al. Rethinking transformer-based set prediction for object detection[C]. Proceedings of the IEEE/CVF international conference on computer vision. 2021: 3611-3620.

[40]Dai Z, Cai B, Lin Y, et al. Up-detr: Unsupervised pre-training for object detection with transformers[C]. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2021: 1601-1610.

[41]Li F, Zeng A, Liu S, et al. Lite detr: An interleaved multi-scale encoder for efficient detr[C]. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2023: 18558-18567.

[42]Sun P, Zhang R, Jiang Y, et al. Sparse r-cnn: End-to-end object detection with learnable proposals[C]. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2021: 14454-14463.

[43]Zhao Y, Lv W, Xu S, et al. Detrs beat yolos on real-time object detection[C]. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2024: 16965-16974.

[44]Nacereddine N, Hamami L, Ziou D. Image thresholding for weld defect extraction in industrial radiographic testing[J]. International Journal of signal processing, 2006, 3(4): 257-265.

[45]Naskath J, Sivakamasundari G, Begum A A S. A study on different deep learning algorithms used in deep neural nets: MLP SOM and DBN[J]. Wireless personal communications, 2023, 128(4): 2913-2936.

[46]Kurani A, Doshi P, Vakharia A, et al. A comprehensive comparative study of artificial neural network (ANN) and support vector machines (SVM) on stock forecasting[J]. Annals of Data Science, 2023, 10(1): 183-208.

[47]Zhang S, Li X, Zong M, et al. Learning k for knn classification[J]. ACM Transactions on Intelligent Systems and Technology (TIST), 2017, 8(3): 1-19.

[48]Askari S. Fuzzy C-Means clustering algorithm for data with unequal cluster sizes and contaminated with noise and outliers: Review and development[J]. Expert Systems with Applications, 2021, 165: 113856.

[49]Chandra M A, Bedi S S. Survey on SVM and their application in image classification[J]. International Journal of Information Technology, 2021, 13(5): 1-11.

[50]Ganie A H, Singh S, Bhatia P K. Some new correlation coefficients of picture fuzzy sets with applications[J]. Neural Computing and Applications, 2020, 32(16): 12609-12625.

[51]Valavanis I, Kosmopoulos D. Multiclass defect detection and classification in weld radiographic images using geometric and texture features[J]. Expert Systems with Applications, 2010, 37(12): 7606-7614.

[52]Alzubaidi L, Zhang J, Humaidi A J, et al. Review of deep learning: concepts, CNN architectures, challenges, applications, future directions[J]. Journal of big Data, 2021, 8: 1-74.

[53] Liu T, Zheng P, Bao J. Deep learning-based welding image recognition: A comprehensive review[J]. Journal of Manufacturing Systems, 2023, 68: 601-625.

[54]García-Pérez A, Gómez-Silva M J, de la Escalera-Hueso A. Improving automatic defect recognition on GDXRay castings dataset by introducing GenAI synthetic training data[J]. NDT & E International, 2025, 151: 103303.

[55]Yang D, Cui Y, Yu Z, et al. Deep learning based steel pipe weld defect detection[J]. Applied Artificial Intelligence, 2021, 35(15): 1237-1249.

[56]Yang L, Fan J, Liu Y, et al. Automatic detection and location of weld beads with deep convolutional neural networks[J]. IEEE Transactions on Instrumentation and Measurement, 2020, 70(8): 1-12.

[57]Ajmi C, Zapata J, Martínez-Álvarez J J, et al. Using deep learning for defect classification on a small weld X-ray image dataset[J]. Journal of Nondestructive Evaluation, 2020, 39(5): 1-13.

[58]Dai W, Li D, Tang D, et al. Deep learning assisted vision inspection of resistance spot welds[J]. Journal of Manufacturing Processes, 2021, 62(9): 262-274.

[59]Ren R, Hung T, Tan K C. A generic deep-learning-based approach for automated surface inspection[J]. IEEE transactions on cybernetics, 2017, 48(3): 929-940.

[60]Dong X, Taylor C J, Cootes T F. Automatic aerospace weld inspection using unsupervised local deep feature learning[J]. Knowledge-Based Systems, 2021, 221: 106892.

[61]Zhang H, Chen Z, Zhang C, et al. Weld defect detection based on deep learning method[C]. 2019 IEEE 15th international conference on automation science and engineering (CASE). IEEE, 2019: 1574-1579.

[62] Yang L, Liu Y, Peng J. An automatic detection and identification method of welded joints based on deep neural network[J]. IEEE Access, 2019, 7: 164952-164961.

[63]Fan K, Peng P, Zhou H, et al. Real-time high-performance laser welding defect detection by combining ACGAN-based data enhancement and multi-model fusion[J]. Sensors, 2021, 21(21): 7304.

[64]Dai W, Li D, Tang D, et al. Deep learning approach for defective spot welds classification using small and class-imbalanced datasets[J]. Neurocomputing, 2022, 477: 46-60.

[65]He X, Luo Z, Li Q, et al. Dg-gan: A high quality defect image generation method for defect detection[J]. Sensors, 2023, 23(13): 5922.

[66]Wang Y, Shi F, Tong X. A welding defect identification approach in X-ray images based on deep convolutional neural networks[C]. Intelligent Computing Methodologies: 15th International Conference, ICIC 2019, Nanchang, China, August 3–6, 2019, Proceedings, Part III 15. Springer International Publishing, 2019: 53-64.

[67]Zou Y, Zhu M, Chen X. A robust detector for automated welding seam tracking system[J]. Journal of Dynamic Systems, Measurement, and Control, 2021, 143(7): 071001.

[68]Liu W, Shan S, Chen H, et al. X-ray weld defect detection based on AF-RCNN[J]. Welding in the World, 2022, 66(6): 1165-1177.

[69]Ji C, Wang H, Li H. Defects detection in weld joints based on visual attention and deep learning[J]. Ndt & E International, 2023, 133: 102764.

[70]Dai X, Chen Y, Yang J, et al. Dynamic detr: End-to-end object detection with dynamic attention[C]. Proceedings of the IEEE/CVF international conference on computer vision. 2021: 2988-2997.

[71]李新越, 刘春秘, 于涵, 等. X射线平板探测器图像噪声分布数学模型构建[J]. 核电子学与探测技术, 2022, 42(06): 1095-1100.

[72]Goodfellow I, Pouget-Abadie J, Mirza M, et al. Generative adversarial networks[J]. Communications of the ACM, 2020, 63(11): 139-144.

[73]He X, Chang Z, Zhang L, et al. A survey of defect detection applications based on generative adversarial networks[J]. IEEE Access, 2022, 10: 113493-113512.

[74]Božič J, Tabernik D, Skočaj D. Mixed supervision for surface-defect detection: From weakly to fully supervised learning[J]. Computers in Industry, 2021, 129: 103459.

[75]Abu-Srhan A, Abushariah M A M, Al-Kadi O S. The effect of loss function on conditional generative adversarial networks[J]. Journal of King Saud University-Computer and Information Sciences, 2022, 34(9): 6977-6988.

[76]Wei Y, Luo X, Hu L, et al. An improved unsupervised representation learning generative adversarial network for remote sensing image scene classification[J]. Remote Sensing Letters, 2020, 11(6): 598-607.

[77]Gao X, Deng F, Yue X. Data augmentation in fault diagnosis based on the Wasserstein generative adversarial network with gradient penalty[J]. Neurocomputing, 2020, 396: 487-494.

[78]Lv G, Israr S M, Qi S. Multi-style unsupervised image synthesis using generative adversarial nets[J]. IEEE Access, 2021, 9: 86025-86036.

[79]Harshvardhan G M, Gourisaria M K, Pandey M, et al. A comprehensive survey and analysis of generative models in machine learning[J]. Computer Science Review, 2020, 38: 100285.

[80]Obukhov A, Krasnyanskiy M. Quality assessment method for GAN based on modified metrics inception score and Fréchet inception distance[C]. Software Engineering Perspectives in Intelligent Systems: Proceedings of 4th Computational Methods in Systems and Software 2020, Vol. 1 4. Springer International Publishing, 2020: 102-114.

[81]Lee J, Lee M. FIDGAN: A generative adversarial network with an inception distance[C]. 2023 International Conference on Artificial Intelligence in Information and Communication (ICAIIC). IEEE, 2023: 397-400.

[82]邓聪, 罗伟坚, 李绪丰. 人工智能技术在射线检测底片评定系统中的应用[J]. 无损检测, 2022, 44(8): 65-68, 73.

[83]Yang L, Wang H, Huo B, et al. An automatic welding defect location algorithm based on deep learning[J]. Ndt & E International, 2021, 120: 102435.

[84]Wu Q, Chen Y, Meng J. DCGAN-based data augmentation for tomato leaf disease identification[J]. IEEE access, 2020, 8: 98716-98728.

[85]Niu Z, Zhong G, Yu H. A review on the attention mechanism of deep learning[J]. Neurocomputing, 2021, 452: 48-62.

[86]Park J, Woo S, Lee J Y, et al. A simple and light-weight attention module for convolutional neural networks[J]. International journal of computer vision, 2020, 128(4): 783-798.

[87]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.

[88]Qin X, Li N, Weng C, et al. Simple attention module based speaker verification with iterative noisy label detection[C]. ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2022: 6722-6726.

[89]Chen Y, Zhang C, Chen B, et al. Accurate leukocyte detection based on deformable-DETR and multi-level feature fusion for aiding diagnosis of blood diseases[J]. Computers in Biology and Medicine, 2024, 170: 107917.

[90]Tian Y, Zhang Y, Zhou D, et al. Triple attention network for video segmentation[J]. Neurocomputing, 2020, 417: 202-211.

[91]Cao L, Wang Q, Luo Y, et al. YOLO-TSL: A lightweight target detection algorithm for UAV infrared images based on Triplet attention and Slim-neck[J]. Infrared Physics & Technology, 2024, 141: 105487.

[92]Juyal P, Kundaliya A. Multilabel image classification using the CNN and DC-CNN model on Pascal VOC 2012 dataset[C]. 2023 International Conference on Sustainable Computing and Smart Systems (ICSCSS). IEEE, 2023: 452-459.