Ge Genbo,Wu Chenchen,Wu Rentao,et al.Real-time monitoring system for petroleum pipeline welding based on spatial perception and improved RLG-YOLO12[J].China Petroleum Machinery,2026,54(9):145-154.
Ge Genbo,Wu Chenchen,Wu Rentao,et al.Real-time monitoring system for petroleum pipeline welding based on spatial perception and improved RLG-YOLO12[J].China Petroleum Machinery,2026,54(9):145-154.DOI: 10.12473/CPM.202601004.
Real-Time Monitoring System for Petroleum Pipeline Welding Based on Spatial Perception and Improved RLG-YOLO12
Welding quality directly determines operational safety and service life of petroleum pipelines. Welding quality evaluation is challenged by diverse morphologies and small scales of defects, and also susceptible to interferences from on-site illumination changes and arc light. This paper presents a
real-time monitoring system for petroleum pipeline welding based on spatial perception and improved RLG-YOLO12. In terms of the model algorithm, RGB images and depth images are fused to construct RGB-D multimodal data to enhance the spatial perception ability for the three-dimensional weld morphology and surface defects; the LAE lightweight adaptive attention mechanism is introduced to strengthen the retention and fusion of detailed information during feature extraction and improve the detection accuracy of tiny and deep defects; and the Ghost Module is used to replace the traditional convolution architecture, thereby significantly reducing the computational complexity and improving the inference speed. In terms of hardware, multiple sensors are integrated for data analysis to realize multi-dimensional monitoring of the welding process and prevent environmental factors unfavorable to welding. Meanwhile, to create a real-time welding scenario for petroleum pipelines, hardware equipment corresponding to the system functions was designed to acquire real-time weld data during welding. The test results show that, on the self-built dataset, the model yields the
R
mAP@50
of 93.2% (an increase of 5.07%), the
R
mAP@50∶95
of 76.8% (an increase of 4.92%), and the computational cost of 16.6 GFLOPs (a decrease of 22.43%). In the actual platform test, the system achieves an accuracy up to 98.4% in the detection on 500 petroleum pipeline samples, and realizes an emergency stop response to serious defects within 100 ms, verifying its stability and practicability in complex industrial scenarios and demonstrating a promising engineering prospect. The research provides an effective intelligent solution for real-time monitoring of welding quality and process optimization.
Cai Xiang , Kang Yihua , Cheng Yangyang . High speed surface detection method and system for steel pipes based on machine vision [J]. China Petroleum Machinery , 2022 , 50 ( 7 ): 143 - 150 .
Yang Xuanyuan , Wang Yihuai . Research on appearance defect detection of welded assembly based on MoblieNetV2 network [J]. Chinese Journal of Sensors and Actuators , 2025 , 38 ( 9 ): 1575 - 1586 .
Tang Wei , Shi Yan , Li Qi , et al . Study on detection method of welding defects in steel based on machine vision and convolutional neural network [J]. Hot Working Technology , 2025 , 54 ( 13 ): 41 - 47 .
Ma Rui , He Deqiang , He Suiqiu , et al . Welding quality detection of metro train based on improved YOLOX [J]. Journal of Railway Science and Engineering , 2023 , 20 ( 10 ): 3998 - 4007 .
Tian Yunjie , Ye Qixiang , Doermann D S . Yolov12: attention-centric real-time object detectors [PP/OL]. arXiv ( 2025-02-18 )[ 2025-04-05 ]. https:∥arxiv.org/abs/2502.12524 https://arxiv.org/abs/2502.12524 .
Arcano-Bea P , Zayas-Gato F , Quintián H , et al . Hybrid vision system for minor pre-assembly identification in a robotic welding cell [C]∥ Hybrid Artificial Intelligent Systems , 2026 : 79 - 90 .
Szőlősi J , Szekeres B J , Magyar P , et al . Welding defect detection with image processing on a custom small dataset: a comparative study [J]. IET Collaborative Intelligent Manufacturing , 2024 , 6 ( 4 ): e70005 .
Wang Guanqiang , Zhang Chizhou , Chen Mingsong , et al . Yolo-MSAPF: multiscale alignment fusion with parallel feature filtering model for high accuracy weld defect detection [J]. IEEE Transactions on Instrumentation and Measurement , 2023 , 72 : 1 - 14 .
Jian Yanying , Liu Tingting , Ni Zhonghua . Welding defect detection with 3D vision and improved YOLOvl1 [J]. Machinery Design & Manufacture , 2026 , ( 6 ): 266 - 273 .
Yu Zhongwen , Guan Qiu , Yang Jianmin , et al . LSM-YOLO: a compact and effective ROI detector for medical detection [C]∥ Neural Information Processing , 2025 : 30 - 44 .
Zhang Zhicong , Pan Feng . LCE-YOLO: a lightweight detection model for aerial images [C]∥ 2025 10th International Conference on Automation, Control and Robotics Engineering (CACRE) , 2025 : 177 - 181 .
Han Kai , Wang Yunhe , Tian Qi , et al . GhostNet: more features from cheap operations [C]∥ 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2020 : 1577 - 1586 .
Guo Xinyan . Research on vehicle detection method based on lightweight convolutional neural network [D]. Nanjing : Nanjing University of Information Science & Technology , 2025 .