RPMI HOME
GaTech HOME
| RPMI HOMEGaTech HOME |
| |
Point-to-Surface Assignment During Registration |
As part of our continuing metrology research program, we have a significant effort underway to develop better and faster ways to measure what we produce, and then to explain those measurements in terms of the fabrication processes that produced the measured parts. There are many technical challenges when developing algorithms to register and analyze three dimensional measurement data. New technologies have become available for quickly generating large data sets, including laser scanners and other optical systems. These technologies provide very dense data sets that can easily contain millions of points. Therefore, these algorithms must be efficient and the computational complexity of the algorithms must be kept low. This is a shift from algorithms that were previously developed to analyze data from touch probe Coordinate Measuring Machines (CMMs) where data sets contain only tens or hundreds of points. The research problem, then, is: How can these large data sets be compared to complex CAD models to yield, in a practical amount of time, information that is useful to manufacturers? This research addresses the analysis problem and contributes to the theoretical body of knowledge for the area. Andre Claudet was the graduate student working on this project, supervised by Tom Kurfess. Generally speaking, we seek to develop algorithms and procedures for extracting artifact quality information from the combination of a set of three-dimensional coordinates with the design CAD model. The specific focus in this project was to solve the problem: Given a point cloud that represents more than one part surface, it is first necessary to determine to which surface each point corresponds. At worst, it is necessary to test each point against each surface in the part. Our research focused on first reducing the number of points to be tested, then on improved methods for eliminating point-surface comparisons and on more efficient point-to-surface deviation calculations. This project has resulted in faster methods for analysis of three dimensional measurement data. Algorithms have been developed and implemented to achieve a reduction in the computational complexity methods currently used at Georgia Tech. The modular framework for the transform allows any first order continuous transform to be included (i.e., fit) in the localization process. |
|
|
 e-mail to : webadmin@rpmi.marc.gatech.edu |
