Article
Three-Dimensional Measurement Method of
Four-View Stereo Vision Based on Gaussian
Process Regression
Miao Gong, Zhijiang Zhang *, Dan Zeng and Tao Peng
Key Laboratory of Specialty Fiber Optics and Optical Access Networks, Joint International Research Laboratory
of Specialty Fiber Optics and Advanced Communication, Shanghai Institute for Advanced Communication and
Data Science, Shanghai University, 99 Shangda Road, Shanghai 200444, China; gongmiaogm@126.com (M.G.);
dzeng@shu.edu.cn (D.Z.); 17820125peng@shu.edu.cn (T.P.)
* Correspondence: zjzhang@shu.edu.cn; Tel.: +86-136-7191-2787
Received: 23 September 2019; Accepted: 14 October 2019; Published: 16 October 2019
Abstract:
Multisensor systems can overcome the limitation of measurement range of single-sensor
systems, but often require complex calibration and data fusion. In this study, a three-dimensional
(3D) measurement method of four-view stereo vision based on Gaussian process (GP) regression is
proposed. Two sets of point cloud data of the measured object are obtained by gray-code phase-shifting
technique. On the basis of the characteristics of the measured object, specific composite kernel functions
are designed to obtain the initial GP model. In view of the difference of noise in each group of
point cloud data, the weight idea is introduced to optimize the GP model, which is the data fusion
based on Bayesian inference method for point cloud data. The proposed method does not require
strict hardware constraints. Simulations for the curve and the high-order surface and experiments
of complex 3D objects have been designed to compare the reconstructing accuracy of the proposed
method and the traditional methods. The results show that the proposed method is superior to the
traditional methods in measurement accuracy and reconstruction effect.
Keywords: multisensor system; Gaussian process regression; Bayesian reasoning method
1. Introduction
The combination of structural illumination and stereo vision has recently provided increased
possibilities for three-dimensional (3D) object measurement, robot vision, and mechanical device
control [
1
–
3
]. Stereo vision measurement methods can be divided into monocular stereo, binocular stereo,
and multivision stereo (MVS) methods [
4
–
6
]. In comparison with the binocular stereo measurement
technology, MVS can enlarge the single measurement range, and the multicameras capture the shape
of the object during the measurement process at a certain moment instantaneously; thus, MVS is
not only suitable for measuring static and nonstrict static objects but also for dynamic and real-time
online measurement.
The measurement plan of four-view stereo measurement method in this study is one of the
applications of MVS technology. For traditional multisensor working independently in its system,
Wu et al. presented a flexible 3D reconstruction method based on phase matching, which reduced
the complexity of calibration between single-sensor systems [
7
]. Xue et al. presented an improved
patch-based multiview stereo method by introducing a photometric discrepancy function based on
a DAISY descriptor; this method obtains good reconstruction results in occlusion and edge regions of
large-scale scenes [
8
]. Zhang et al. created a multiview stereo vision system for true 3D reconstruction,
modeling, and phenotyping of plants. This system yielded satisfactory 3D reconstruction results
and demonstrated the capability to study plant development where the same plants were repeatedly
Sensors 2019, 19, 4486; doi:10.3390/s19204486 www.mdpi.com/journal/sensors
