Real-time detection method for weld feature of construction site steel structure based on top-down paradigm

被引：0
作者：
Cheng J. [1 ,2 ]
Jin H. [1 ,2 ]
Zheng Z. [1 ,2 ]
Jiang L. [1 ,2 ]
Luo Q. [3 ]
Dong K. [3 ]
Zhou J. [4 ]
Chen X. [4 ]
机构：
[1] School of Civil Engineering, Southeast University, Nanjing
[2] Jiangsu Key Laboratory of Engineering Mechanics, Southeast University, Nanjing
[3] China Construction Science and Industry Jiangsu Corporation Ltd., Nanjing
[4] China Construction Steel Structure Jiangsu Corporation Ltd., Taizhou
来源：
Dongnan Daxue Xuebao (Ziran Kexue Ban)/Journal of Southeast University (Natural Science Edition) | 2023年 / 53卷 / 06期
关键词：
deep learning; human pose estimation; intelligent construction; structural light vision; weld feature point;
D O I：
10.3969/j.issn.1001-0505.2023.06.017
中图分类号：
学科分类号：
摘要：
Aiming at the problems of complex background textures in on-site weld images and difficulty in locating profile feature points, a real-time detection method for end-to-end weld features is proposed to improve the welding efficiency and quality of building steel structures. Adopting the idea from human pose estimation, the extraction of weld feature points is equivalent to the key point detection task of the human skeleton. Following the top-down paradigm of pose estimation, the extraction method for weld feature of building steel structure in construction site is established. The RTMdet object detector is first introduced to quickly locate the welding profile area, then the RTMPose pose estimation model is applied to detect profile feature points in the target region. By converting the feature point coordinate regression localization into a classification problem of horizontal and vertical coordinates, the localization accuracy and efficiency is effectively improved. Experimental results demonstrate that compared to the welding identification methods based on digital image processing or regression with fully connected layers, the method can rapidly and precisely extract the welding feature points from images containing complex information. The localization error of the feature point in a single image is less than 2 pixels, and the average processing time is 38. 2 ms, meeting the requirements for automated welding on construction sites. © 2023 Southeast University. All rights reserved.
引用
收藏
页码：1100 / 1110
页数：10
相关论文
共 18 条
  • [1] Cheng J M, Jin H, Zheng Z J, Et al., Welding seam groove recognition of steel structure on construction site based on machine vision [J], Journal of Southeast University (Natural Science Edition), 53, 1, pp. 86-93, (2023)
  • [2] Huang Y G, Wang D Q, Jiang M, Et al., Laser fringe segmentation and feature points location method of weld image based on multi-task learning [J], Chinese Journal of Lasers, 50, 16, (2023)
  • [3] Deng L Y, Lei T, Wu C Q, Et al., A weld seam feature real-time extraction method of three typical welds based on target detection [J], Measurement, 207, (2023)
  • [4] Zhang Y S, Yang G W, Wang Q Q, Et al., Weld feature extraction based on fully convolutional networks [J], Chinese Journal of leasers, 46, 3, pp. 36-43, (2019)
  • [5] Huang S J, Xu Y L, Yang X J, Et al., ROI weld feature extraction algorithm of real-time tracking, Journal of Shanghai Jiao Tong University, 50, 12, pp. 1877-1880, (2016)
  • [6] Yang X J, Xu Y L, Huang S J, Et al., A recognition algorithm for feature points of V-groove welds based on structured light [1], Journal of Shanghai Jiaotong University, 50, 10, pp. 1573-1577, (2016)
  • [7] Li Y Y, Zhang Z Y, Yuan L., Survey on linear structured light stripe center extraction [1], Laser & Optoelectronics Progress, 50, 10, pp. 13-22, (2013)
  • [8] Chen B, He S, Liu 1, Et al., Weld structured light image segmentation based on lightweight DeepLab v3 + Net-work, Chinese Journal of Lasers, 50, 8, pp. 49-58, (2023)
  • [9] Zhang S K, Wu Q X, Lin Z Y., Detection and segmentation of structured light stripe in weld image [J], Acta Optica Sinica, 41, 5, pp. 88-96, (2021)
  • [10] Zou Y B, Wei X Z, Chen I X, Et al., A high-accuracy and robust seam tracking system based on adversarial learning, IEEE Transactions on Instrumentation and Measurement, 71, pp. 1-13, (2022)
12