Feature-Based Three-Dimensional Registration for Repetitive Geometry in Machine Vision

Abstract

As an important step in three-dimensional (3D) machine vision, 3D registration is a process of aligning two or multiple 3D point clouds that are collected from different perspectives together into a complete one. The most popular approach to register point clouds is to minimize the difference between these point clouds iteratively by Iterative Closest Point (ICP) algorithm. However, ICP does not work well for repetitive geometries. To solve this problem, a feature-based 3D registration algorithm is proposed to align the point clouds that are generated by vision-based 3D reconstruction. By utilizing texture information of the object and the robustness of image features, 3D correspondences can be retrieved so that the 3D registration of two point clouds is to solve a rigid transformation. The comparison of our method and different ICP algorithms demonstrates that our proposed algorithm is more accurate, efficient and robust for repetitive geometry registration. Moreover, this method can also be used to solve high depth uncertainty problem caused by little camera baseline in vision-based 3D reconstruction.

Three-dimensional measurement of small inner surface profiles using feature-based 3-D panoramic registration

“…As the gold standard of point cloud registration, ICP could not be used to build the multiple quadrants into a complete model due to the repetitive geometrical nature of the spiral thread. 35 The existing outliers in the registered model were then trimmed by a statistical filter based on their spatial distribution. The comparison among the final reconstructed point cloud and x-ray CT data shows that x-ray CT generates a much denser point cloud than our vision-based 3-D reconstruction approach, but it also comes with a parameter (iso) that greatly affects the measurement.…”

“…To register two point clouds together, a feature-based 3-D registration algorithm was introduced in our previous work. 35 Building on this work, a more advanced algorithm specific to the inspection of a threaded hole is proposed. This algorithm solves the 3-D registration problem of multiple point clouds that cover a 360 deg perspective of an object, which we call feature-based 3-D panoramic registration.…”

“…However, ICP would fail in our case, resulting in the complete overlap of different quadrants due to their highly similar geometry. 35 In this paper, a feature-based 3-D panoramic registration algorithm is proposed to achieve the registration of these highly similar quadrants from 360 deg scanning into an accurate 3-D surface reconstruction by taking advantage of the robustness of image features and their 3-D correspondences.…”

“…To register all quadrants together, the standard approach using ICP did not work well for this case due to the highly similar geometry of these quadrants. 35 Our proposed feature-based 3-D panoramic registration was applied to take advantage of the high accuracy, robustness, and distinguish-ability of SIFT features. By applying the same feature detection and matching algorithm on the angular image sequences of two neighbor quadrants ( i ’th and j ’th), the common features ( P i,j and P j,i ) were detected.…”

Advanced teleoperation and control system for industrial robots based on augmented virtuality and haptic feedback

“…In particular, a network of three 3D sensors (Kinect cameras) with eye-tohand configuration is used in this work. The process of "registration" is needed to obtain the transformation relating two views of the same workpiece [38]. This process is useful to obtain the position and orientation (i.e., the pose) of a workpiece by matching the point cloud acquired by the camera with the virtual model of the workpiece.…”