TY - JOUR
T1 - Enhanced point cloud registration for workpieces using triangular constraint sampling consistency in complex industrial environment
AU - Guo, Zhentao
AU - Ma, Hongbin
N1 - Publisher Copyright:
© The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2026.
PY - 2026/3
Y1 - 2026/3
N2 - Accurately identifying and locating workpieces in complex industrial environment poses significant challenges due to high noise levels and large point cloud datasets. This paper proposes a method for acquiring workpiece point clouds using 3D structured light measurement for generating high-quality and reliable workpiece datasets. The proposed approach combines noise filtering, adaptive downsampling, PCA-FPFH feature extraction, and TCSAC (triangular constraint sampling consistency) to address the challenges posed by excessive noise and the substantial size of point cloud datasets, which significantly increase the computational overhead of registration algorithms. Additionally, the proposed method addresses the issue where most workpieces in industrial environment have similar features, which reduces the matching accuracy of point cloud registration. Experimental results demonstrate that our method improves both the accuracy and speed of point cloud registration in complex industrial environment and demonstrates strong performance on the 3DMatch, 3DLoMatch, and odometryKITTI open-source benchmarks. The relevant code is available at https://github.com/ZhentaoGuo/TCSAC.
AB - Accurately identifying and locating workpieces in complex industrial environment poses significant challenges due to high noise levels and large point cloud datasets. This paper proposes a method for acquiring workpiece point clouds using 3D structured light measurement for generating high-quality and reliable workpiece datasets. The proposed approach combines noise filtering, adaptive downsampling, PCA-FPFH feature extraction, and TCSAC (triangular constraint sampling consistency) to address the challenges posed by excessive noise and the substantial size of point cloud datasets, which significantly increase the computational overhead of registration algorithms. Additionally, the proposed method addresses the issue where most workpieces in industrial environment have similar features, which reduces the matching accuracy of point cloud registration. Experimental results demonstrate that our method improves both the accuracy and speed of point cloud registration in complex industrial environment and demonstrates strong performance on the 3DMatch, 3DLoMatch, and odometryKITTI open-source benchmarks. The relevant code is available at https://github.com/ZhentaoGuo/TCSAC.
KW - Adaptive downsampling
KW - Industrial environment
KW - PCA-FPFH feature extraction
KW - Point cloud registration
KW - TCSAC
UR - https://www.scopus.com/pages/publications/105035008485
U2 - 10.1007/s00371-026-04421-1
DO - 10.1007/s00371-026-04421-1
M3 - Article
AN - SCOPUS:105035008485
SN - 0178-2789
VL - 42
JO - Visual Computer
JF - Visual Computer
IS - 5
M1 - 221
ER -