Real-time detection of surface deformation and strain in recycled aggregate concrete-filled steel tubular columns via four-ocular vision

• Published in: Robotics and computer-integrated manufacturing Vol. 59; pp. 36 - 46
• Main Authors: Tang, Yunchao, Li, Lijuan, Wang, Chenglin, Chen, Mingyou, Feng, Wenxian, Zou, Xiangjun, Huang, Kuangyu
• Format: Journal Article
• Language: English
• Published: Oxford Elsevier Ltd 01.10.2019; Elsevier BV
• Subjects: Concrete aggregates; Cyclic loads; Deformation; Image reconstruction; Laser range finders; Point cloud registration; Real time; Real-time detection; Recycled materials; Remanufacturing; Stereo matching; Stereo vision; Stereoscopic vision; Strain; Three dimensional models; Vision systems; Point cloud registration; Stereo vision; Real-time detection; Remanufacturing; Stereo matching
• ISSN: 0736-5845; 1879-2537
• DOI: 10.1016/j.rcim.2019.03.001

•A mark-free vision-based dynamic real-time measurement technique is proposed for full field strain and 3D deformation detection.•Point cloud combination and 3D reconstruction are applied.•The non-contact method results truly reflect the deformation process and full-field strain values.•The standard deviation mean value measured by the proposed method relative to the true value is 1.23%.•The proposed method can be used as a reasonable alternative to traditional contact methods. This study presents a dynamic real-time detection method for surface deformation and full field strain in recycled aggregate concrete-filled steel tubular columns (RACSTCs). Automatic calibration and dynamic surface tracking measurement are utilized and mathematical models combining the four-ocular visual coordinates and point cloud matching are proposed. The 3D deformation of the RACSTCs under low cyclic loading are gathered every 60 s as sample images. The four-ocular vision system constitutes two groups of binocular stereoscopic vision systems. The 3D deformation surface is reconstructed by multi-ocular vision coordinate association, image preprocessing, and point cloud registration. The measurement results are validated by comparison against laser range finder data. The standard deviation mean value of 10 specimens measured by the proposed method relative to the true value is 1.23%; the mean error of the maximum absolute value is 2.82%.