STL file generation from measured point data by segmentation and Delaunay triangulation
Introduction
Since the life cycle of products is getting shorter due to the rapid industrial development, the efficient reduction of the new product development time and the manufacturing process should be the significant issue. To cope with customers' diverse needs and aesthetic design, which are considered in design process, reverse engineering has become important these days in real workshops. And some industrial parts, which do not have CAD data, still exist and, when these parts need to be copied, reproduced or remodeled for design change, reverse engineering gets useful.
Typical steps of reverse engineering include the scanning of a clay or wood model and the generation of manufacturing data which can be used in CAD/CAM systems. For the implementation of reverse engineering, fast measuring speed and precision are required to the measuring machine, and a laser scanner is currently used a lot satisfying both advantages. And the line-typed and cross-sectional data make it easy to model a surface from the measured data. But some drawbacks come from too many measured points, which are obstacles for surface generation such as sampling of points, segmentation, and many control point's problems. Direct generation of STL file from the scanned data is favorable in that it can reduce the time and error in modeling process.
This paper proposes an efficient approach to reduce the amount of data while generating STL file directly from the measured point data with maintaining their precision as shown in Fig. 1. Both segmentation and Delaunay triangulation are considered to provide the best tradeoff between reduction rate and accuracy.
Lots of research has been focused on shape reconstruction from point data, but there are some limitations to adapt to the measured data with noise as follows.
Noise is globally distributed in the scanned data from laser scanning, especially in the round surface, surface fillet, and smooth boundary region. It is hard to find definitely the segment boundaries around these surfaces even in the human eye. Smooth filtering can decrease severe noise, but it is hard to remove the unfiltered noise, which is globally distributed. And this noise is included in the resultant model. Thus general methods, which are based on the local curvature between points or the error between the point data and the generated model, are not effective in that the candidate regions for removal are not widely selected and the repetition of the methods concentrates on several regions.
If previous mesh simplification methods, which, in general, starts with a polygonization and successively simplify it, are used to the data with noise, the region with noise is distinctive from the modified region and the boundaries and geometric characteristics are not so clearly maintained.
The boundaries of the data from CAD systems are definitely found considering only the angle between triangles, but the satisfactory boundaries of the data from laser scanning are not computed due to the noise globally distributed.
The segment boundaries, proposed in this paper considering the relation of a triangle and its adjacent triangles, make it possible to remove noise a bit and maintain geometric characteristics which are close to the original feature. The retriangulation is done in segment boundaries, the characteristic lines, and this means the regions for retriangulation are widely distributed under the same removal ratio.
Previous research on segmentation has been applied to the data with good quality. That is to say, many examples have used the data modeled in CAD systems, tessellated data or cross-sectional data of them. Simple objects with sharp edges are used in some experimental results for easy segmentation. Some data begin with the assumption that noise should be totally removed. Segmentation is hard to apply to the scanned data of an object with scratches and grooves, which is worn out for a long time.
Scanned data can be reduced by Delaunay triangulation with relatively high precision. But repetitive implementation of the triangulation may remove the geometric characteristics especially in the region such as round surface, surface fillet and smooth boundary region. Delaunay triangulation is performed maintaining the segment boundaries so that the scanned data are reduced with maintaining its original geometric characteristics.
Section snippets
Related research
Most of researches in reverse engineering have been focused on the development of a measuring machine, surface generation, and the improvement of surface quality.
Data acquisition [1] in reverse engineering by a contact and non-contact measuring machine is compared and the real problems during measurement are examined. And some issues on creating B-rep models are described through characterization of geometric models, segmentation and surface fitting.
Previous researches on segmentation are
Generation of triangular net
Scanned data from laser scanning have the arranged form either in planar or circular type. A triangular net can be generated from the data in planar type by repeated connection of a point of one polyline to that of another polyline using the Christiansen algorithm [20].
But, in the case of the data arranged in circular type, some problems may arise while not considering the topological relationships among 3D points as shown in Fig. 2. These problems happen when the shapes of two loops are
Segmentation of measured data
Segmentation is defined as the division of measured points in several regions to generate smooth surface in each region which is suitable for computer operation.
Segmentation only based on the angle between triangles is valid to the STL file which is obtained from the tessellation of the data which are modeled in CAD systems. But, the scanned data with some noise are not clear enough depending on machining and surface quality as shown in Fig. 5, and the segmentation of this kind of data based on
Data reduction by Delaunay triangulation
Delaunay triangulation is done by forming a circumscribed circle from three arbitrary points and letting any other point not to exist in that circle. A triangle with small and narrow angle should be avoided for STL generation and real fabrication by RP equipments, and it makes the smallest angle of all triangles as large one.
For reduction of point data, select and remove a vertex which will make an error less than tolerance when removed. The region around the vertex is retriangulated by
Experimental results
Some experiments are performed to demonstrate the theories in this paper with the software developed using Visual C++ 6.0 and OpenGL. A triangular net is generated from the scanned data, and segmentation and Delaunay triangulation are performed to two models. One is a mask model for planar scanning with scanned data of 20,337 points and the other is a faucet model for circular scanning with scanned data of 24,138 points.
The criteria for the segmentation of the mask model is that the joining
Conclusion
In this paper, the more systematic approach to generate STL file by segmentation and Delaunay triangulation from the scanned data is presented.
A triangular net is generated to maximize the smallest angle over all triangulations by adopting a bounding box and max–min angle criterion with the consideration of topological relationships among 3D points. Thus it helps to improve the surface quality of RP parts, the reduction of time and labor in postprocess.
Segmentation is performed based on local
Seok-Hee Lee is a professor in the School of Mechanical Engineering at the Pusan National University. He received BS degree from Seoul National University in 1976, MS degree from Korea Advanced Institute of Science and Technology in 1978, and PhD from University of Manchester Institute of Science and Technology in 1985, all in Mechanical Engineering. He was a visiting professor at the University of Wisconsin, Madison in 1994. His current research interests include production system engineering,
References (23)
- et al.
Reverse engineering of geometric models-an introduction
Comput Aided Des
(1997) - et al.
Segmentation of a wrap-around model using an active contour
Comput Aided Des
(1997) - et al.
Reverse engineering modeling of free-form surfaces from point clouds subject to boundary conditions
J Mater Process Technol
(1998) - et al.
Segmentation of measured point data using a parametric quadric surface approximation
Comput Aided Des
(1999) - et al.
Smooth-surface approximation and reverse engineering
Comput Aided Des
(1991) - et al.
A new approach to reverse engineering based on vision information
Int J Mach Tools Manufact
(1998) A data reduction scheme for triangulated surfaces
Comput Aided Geometric Des
(1994)- et al.
Mesh simplification with smooth surface reconstruction
Comput Aided Des
(1998) - et al.
Algorithm for approximate NURBS skinning
Comput Aided Des
(1996) - et al.
Reverse engineering of free-form surfaces
J Mater Process Technol
(1998)
An adaptive method for smooth surface approximation to scattered 3D points
Comput Aided Des
Cited by (37)
Iterative offset-based method for reconstructing a mesh model from the point cloud of a pig
2022, Computers and Electronics in AgricultureCitation Excerpt :An alternative approach for overcoming this problem is to reconstruct 3D models (mesh models) from the point clouds obtained from livestock to calculate their volume, and based on these results, estimate their weight. Lee et al. (2002) attempted to create a stereolithography (STL) file, which is a mesh type technique, using point clouds obtained by a laser scanner. However, this method of using Delaunay triangulation was limited as it required a significant amount of calculation resources to create a mesh model (Lee et al., 2001).
Tool path generation, for complex surface machining, using point cloud data
2015, Procedia CIRPIntegrating cross-sectional imaging based reverse engineering with rapid prototyping
2006, Computers in IndustryCitation Excerpt :The region to be triangulated is selected by the deviation of a triangle and its surrounding ones, and the triangulation is implemented based on the distance between points of the region. In the case of laser scanning, Lee et al. [10] proposed an approach to generate STL file from measurement point data by segmentation and Delaunay triangulation. Segmentation is performed based on local and global curvatures between triangles, and segments are classified as plane, smooth, and rough regions.
Reduction of post-processing for stereolithography systems by fabrication-direction optimization
2005, CAD Computer Aided DesignA Critical Review of Construction Using 3D Printing Technology
2023, Journal of Engineering for Sustainable Buildings and CitiesSTL Generation from Point Cloud Data with User-Controlled Triangulation for Additive Manufacturing
2022, Lecture Notes in Mechanical Engineering
Seok-Hee Lee is a professor in the School of Mechanical Engineering at the Pusan National University. He received BS degree from Seoul National University in 1976, MS degree from Korea Advanced Institute of Science and Technology in 1978, and PhD from University of Manchester Institute of Science and Technology in 1985, all in Mechanical Engineering. He was a visiting professor at the University of Wisconsin, Madison in 1994. His current research interests include production system engineering, CAD/CAM, RP and reverse engineering.
Ho-Chan Kim is a PhD candidate in the Department of Mechanical and Intelligent Systems Engineering at the Pusan National University. He received his BS in the Department of Production and Mechanical Engineering from the Pusan National University in 1996 and MS in the Department of Mechanical and Intelligent Systems Engineering from the Pusan National University in 1998. He worked at Samsung electro-mechanics as a research engineer. His research interests include CAD/CAM, RP and reverse engineering.
Sung-Min Hur is a PhD candidate in the Department of Mechanical and Intelligent Systems Engineering at the Pusan National University. He received his BS in the Department of Production and Mechanical Engineering from the Pusan National University in 1996 and MS in the Department of Mechanical and Intelligent Systems Engineering from the Pusan National University in 1998. His research interests include CAD/CAM, RP and reverse engineering.
Dong-Yol Yang is a professor in the Department of Mechanical Engineering at the Korea Advanced Institute of Science and Technology. He received BS degree from Seoul National University in 1973, MS degree from Korea Advanced Institute of Science and Technology in 1975, and PhD from Korea Advanced Institute of Science and Technology in 1978, all in Mechanical Engineering. His current research interests include net shape manufacturing, plasticity, materials forming and RP.