Open AccessA Dense 3-D Point Cloud Measurement Based on 1-D Background-Normalized Fourier Transform
Author(s) -
Ruiying Liao,
Linghui Yang,
Luyao Ma,
Jincheng Yang,
Jigui Zhu
Publication year - 2021
Publication title -
ieee transactions on instrumentation and measurement
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.82
H-Index - 119
eISSN - 1557-9662
pISSN - 0018-9456
DOI - 10.1109/tim.2021.3075740
Subject(s) - power, energy and industry applications , components, circuits, devices and systems
The 3-D surface measurement plays a significant role in the industrial area. With the advantages of high resolution and acquisition rate, dual-line-scan camera systems have been gradually studied for dynamic 3-D measurement. However, the partial overexposure and the limited measurement depth range of dual-line-scan cameras are the two main elements affecting the quality of the point cloud. Both of these problems are caused by incorrect pixel matching. Therefore, this article presents a matching strategy for dual-line-scan cameras based on the 1-D background-normalized Fourier transform (1DBNFT) to expand the depth range of measurement and deal with the partial overexposure. The wrapped phases extracted by 1DBNFT, the unwrapped phase extracted by projection distance minimization (PDM), and the matching principle are described in detail. We also analyze the possible errors in the system from three aspects: noncoplanar error, installation error, and motion error. Finally, the comparative experiments and accuracy verification experiments demonstrate the effectiveness of our algorithm.