Research on the Application of YOLOv8 Improvement Strategy in Welding External Defect Detection Model
(Jingdezhen Ceramic University, Jingdezhen 333403, China)
Abstract
Aiming at the challenges of missed detection, false detection and arbitrary distribution of defect angles in small target defect detection in welding external defect image detection, an optimized and improved model based on YOLOv8, YOLO-weld, is proposed. Firstly, SPPF-weld and C2f_DBB_CBAM modules are designed to enhance the context information aggregation ability and multi-scale feature fusion effect of YOLO-weld model. Secondly, the OBB detection head is introduced to accurately capture directional defects, so as to improve the detection accuracy. Finally, the experimental results show that compared with the YOLOv8 model, the YOLO-weld model improves the accuracy, recall, harmonic mean (F1) and mAP@0.5 by 3.2%, 2.6%, 4.0% and 3.9%, respectively, which fully proves the effectiveness of the YOLO-weld model improvement.
Keywords
welding external defect identification; non-destructive testing; YOLOv8; C2f_DBB
References
- [1] KERMORGANT O. A Magnetic Climbing Robot to Perform Autonomous Welding in the Shipbuilding Industry [J].Robotics and Computer-Integrated Manufacturing,2018,53:178-186.
- [2] MHAMED I B,ABID S,FNAIECH F. Weld Defect Detection Using a Modified Anisotropic Diffusion Model [J].EURASIP Journal on Advances in Signal Processing,2012,2012:1-12.
- [3] DAS B,PAL S,BAG S. A Combined Wavelet Packet and Hilbert-Huang Transform for Defect Detection and Modelling of Weld Strength in Friction Stir Welding Process [J].Journal of Manufacturing Processes,2016,22:260-268.
- [4] GUO R Y,LIU H,XIE G,et al. Weld Defect Detection From Imbalanced Radiographic Images Based on Contrast Enhancement Conditional Generative Adversarial Network and Transfer Learning [J].IEEE Sensors Journal,2021,21(9):10844-10853.
- [5] ZHANG K H,SHEN H K. Solder Joint Defect Detection in the Connectors Using Improved Faster-RCNN Algorithm [J/OL].Applied Sciences,2021,11(2):576[2025-01-03].https://doi.org/10.3390/app11020576.
- [6] LIU J X,LI K X. Intelligent Metal Welding Defect Detection Model on Improved FAST-PNN [J/OL].Coatings,2022,12(10):1523[2025-01-10].https://doi.org/10.3390/coatings12101523.
- [7] KIM K,KIM K S,PARK H J. Multi-Branch Deep Fusion Network-Based Automatic Detection of Weld Defects Using Non-Destructive Ultrasonic Test [J].IEEE Access,2023,11:114489-114496.
- [8] ZHANG S W,WANG X Q,ZHANG T. Combining Multi-Scale U-Net With Transformer for Welding Defect Detection of Oil/Gas Pipeline [J].IEEE Access,2024,13:5437-5445.
- [9] LI Y F,GAO P Y,LUO Y B,et al. Automatic Detection and Classification of Natural Weld Defects Using Alternating Magneto-Optical Imaging and ResNet50 [J/OL].Sensors,2024,24(23):7649[2025-01-17].https://doi.org/10.3390/s24237649.
- [10] ZHANG G,WANG C T,XIAO D Q. A Novel Daily Behavior Recognition Model for Cage-reared Ducks by Improving SPPF and C3 of YOLOv5s [J/OL].Computers and Electronics in Agriculture,2024,227:109580[2024-12-23].https://doi.org/10.1016/j.compag.2024.109580.
- [11] JOOSHIN H K,NANGIR M,SEYEDARABI H. Inception-YOLO: Computational Cost and Accuracy Improvement of the YOLOv5 Model Based on Employing Modified CSP, SPPF, and Inception Modules [J].IET Image Processing,2024,18(8):1985-1999.
- [12] LUO Z,WANG C,QI Z,et al. LA_YOLOv8s: A Lightweight-Attention YOLOv8s for Oil Leakage Detection in Power Transformers [J].Alexandria Engineering Journal,2024,92:82-91.
- [13] Zou Zhentao Improvement of YOLO Algorithm for Remote Sensing Image Object Detection [D]. Guiyang: Guizhou University, 2024
About the authors
Zhu Yonghong (1965-), male, Han ethnicity, from Poyang, Jiangxi Province, professor, PhD, research interests: intelligent detection and control, image processing, information fusion, and network control; Yang Kaifu (1997-), male, Han ethnicity, from Yichang, Hubei, is currently pursuing a master's degree in intelligent detection and control, image processing, and information fusion.