Top row: The Iterative Closest Point method aligns a point cloud model (red) with a reference model (green) to determine the position of a rigid object. Bottom row: A random sampling of points, an interest sampling of points, and the test scene.

Augmented reality relies on object tracking which means in this context to be able to identify a physical object and to

follow its position and orientation with respect to a camera. One method for this purpose is Iterative Closest Points (ICP). ICP uses a point could from the environment and matches this object with reference models of the object to track. It solves a least square problem to align the environment model with the reference model and to minimize the root mean square distance.


The goal of this project is to facilitate the tracking of multiple physical objects with ICP with camera frame rates and to increase the robustness to track objects of mechanical engineering on the shop floor.


The domain that is addressed in this research is mechanical engineering and assembly assistance in particular. The assembly of a mechanical part typically incorporates several parts that need to be tracked. In our research, we develop methods that allow us to determine several objects. We follow two approaches. First, we combine all reference objects into one overall reference object. Instead of matching multiple objects subsequently, we match one reference object that includes all single objects. Since the performance of ICP also depends on the total number of points, we reduce the number of reference points after those have been merged.


Second, we investigate different point sampling algorithm, which selects a subset of points from the environment point cloud. Several methods are under investigation: uniform sampling, which selects points in a uniform grid distribution; random sampling, which selects random points; interest point sampling, a method we introduced, selects points around an interest point. The points are selected using a Gaussian distribution. Points that are close to the interest point are selected with a higher probability than points that are far away. The ICP-based tracking method has been implemented and tested. It allows us to identify and track physical objects as well as superimpose these objects with a virtual 3D model.


For further information:

Garrett, T., Debernardis, S., Radkowski, R., Chang, C.K., Fiorentino, M., Uva, A., and Oliver, J., 2014, "Rigid Object Tracking Algorithms for low-cost Augmented Reality Devices," Proc. of the ASME 2014 International Design Engineering Technical Conferences & Computers and Information in Engineering Conference, IDETC/CIE 2014, August 17-20, 2014, Buffalo, NY, USA

About Us


Recent Works

The Augmented Reality Lab explores the augmented reality (AR) technology and its capabilities for engineering applications.

+1 (515) 294-7044

Iowa State University

1620 Howe Hall

Ames, IA 50011-2161

© 2014 All Rights Reserved