| Summary: | 碩士 === 國立中正大學 === 機械系 === 92 === In order to obtain a complete three-dimensional data of a model for reverse engineering, we need to scan a model two or more times. However, with improvements in scanners and accompanying software, problems related to scanning a model and to the techniques used for processing a large amount of scan data have occurred.
Our study developed and applied techniques for processing a large amount of point cloud data for use in reverse engineering. These techniques can be divided into two major phases: registration and integration. For the registration phase, we improved the precision registration algorithm (ICP). This improvement made the ICP steadier and speeded up its performance. For the integration phase, we adopted the digitation (voxel) integration algorithm. This algorithm can be combined with the fill hole algorithm which is used at a later time. After the two phases are completed, we get a complete digital model. In addition to the proposals for improving the ICP and adopting the digitation integration algorithm, we also propose a brand new automatic initial registration algorithm which can automate the registration and integration processes as well as minimize errors. The method proposed in this thesis was compared with other business software on the market. Our system is capable of processing more scanning data than the business software systems compared using the same amount of memory.
|
|---|
