Session: 06-07-01- Sustainable design
Paper Number: 90352
90352 - Automated Digitization of 3d Structures Using AprilTag and SG-Library in MATLAB
In the course of Industry 4.0, digitalization is playing an increasing role in all engineering applications. In the design and manufacturing process reverse engineering has been broadly used to produce a digital 3D replica model of existing mechanical parts. This can be useful both in the industrial environment and in research and development, since 3D models are often not available for old mechanical parts and prototypes. The created model can then be documented and modified for further development.
Conventional techniques for the creation of digital 3D models can be divided into contact and non-contact measurement methods. In contact measurement methods, contact probes are typically mounted on the end of a measurement arm that physically touches the measuring object to acquire coordinate points for geometry creation. Non-contact measurement methods include different 3D scanning technologies such as LASER triangulation, structured light and photogrammetry. Industrial computed tomography (ICT) is another technique for reverse engineering, which does not require disassembly. While contact-based coordinate measurement arms can achieve high accuracy, the measuring device must be mounted at a fixed point, thus limiting the measuring volume. The measuring volume of ICT devices is also a limiting factor. 3D scanning technologies usually have a larger measuring volume. However, the 3D models created using these techniques are reconstructed based on point cloud, which leads to rough surface quality. In addition, the above-mentioned methods, especially the contact-based coordinate measurement arms and ICT, are associated with high costs, which hinders wide distribution for home use.
This work proposes a novel low-cost method for automated digitalization of 3D structures based on AprilTag and SG-Library. AprilTag is a visual fiducial system, in which the fiducial markers can be printed with ordinary paper printers. The AprilTag detection algorithm computes the precise 3D position and orientation of AprilTags relative to the camera. The measuring systems consists of two major parts: 1) a coordinate measuring system using AprilTag detection and 2) a probe with AprilTags attached for surface measuring of mechanical parts. Since the relative pose (3D position and orientation) from the tip of the probe to the origin of the AprilTag coordinate system is known, the relative pose from the tip to camera coordinate system can be derived. The measured pose data is then imported into MATLAB for further geometry generation using Solid Geometry (SG-) Library, which is a surface-modelling toolbox for automated design process in MATLAB. In cases where the geometry has a fixed cross-section, and this cross-section is already known, only points for the extrusion path will be acquired. The extrusion is carried out automatically using SG-Library. Otherwise, the geometry creation starts with the creation of 2D contours, similar to a conventional CAD modelling process. For each type of 2D contour, a set of data points is acquired automatically using the probe and coordinate measuring system (e.g., 3 points for circles or triangles and 4 points for quadrilaterals). Based on 2D contours, 3D structures are generated. Special 3D structures such as pyramids or spheres can also be created directly. One major advantage compared to other conventional 3D scanning technologies is that the proposed method avoids redundant triangulation and ensures that the created 3D model has good surface quality. The measurement volume can also be adjusted by changing the camera lenses. Furthermore, the proposed method can be implemented without high cost using standard single board computers and camera modules.
To verify the proposed method, the automated process measures and creates prototypes based on aluminum construction profiles and samples of 3D structures such as cylinders, cubs, spheres and tetrahedrons.
In summary, this paper proposes a novel low-cost measurement method based on pose estimation using AprilTag for automated digitalization of 3D structures. Automatic geometry generation is achieved using SG-Library in MATLAB. More interface functions converting acquired 3D pose data into geometry can be implemented to increase the functionality of this system.
Presenting Author: Dingzhi Zhang Institute of Micro Technology and Medical Device Technology, Technical University of Munich
Presenting Author Biography: Dingzhi Zhang is a PhD student at the Institute of Micro Technology and Medical Device Technology at TUM, Germany. His research focus on vision-based navigation and classification for robotic applications and automated design process. Dingzhi Zhang received his Bachelor’s degree (2017) in Mechanical Engineering and Master's degree (2020) in Medical Technology and Engineering from the Technical University of Munich (TUM), Germany. During his studies towards his Master’s degree, he stayed at the KTH Royal Institute of Technology in Stockholm, Sweden, for an exchange study in Medical Informatics.
Authors:
Dingzhi Zhang Institute of Micro Technology and Medical Device Technology, Technical University of MunichFelix Pancheri Institute of Micro Technology and Medical Device Technology, Technical University of Munich
Christoph Rehekampff Institute of Micro Technology and Medical Device Technology, Technical University of Munich
Yilun Sun Institute of Micro Technology and Medical Device Technology, Technical University of Munich
Tim C. Lueth Institute of Micro Technology and Medical Device Technology, Technical University of Munich
Automated Digitization of 3d Structures Using AprilTag and SG-Library in MATLAB
Paper Type
Technical Paper Publication