Research On Data Preprocessing Algorithms In Reverse Engineering And Its Software Implementation

Reverse Engineering(RE) is a way to convert physical models of products intodigital models, and it can greatly shorten the development cycle of products. Thus, ithas been widely applied in many areas. Data preprocessing is an important step ofreverse engineering as it affects the quality of subsequent model reconstruction. Inthis thesis, some key technologies in data preprocessing are investigated.Firstly, denoising methods for point cloud are investigated. The cause of noise andexistingdenoising methodsare analysed. Then we introduce an algorithm based on thek-d tree with consideration of the characteristics of scattered point cloud. The result isquite good in real processingusing this algorithm.Then,simplification methods for point cloud are investigated. Several existingmethods, including curvature based samplingmethod and bounding-box basedmethod,are analysed. Then we introduce an algorithm using the distance between thepoint and the tangent plane fitted by the neighbour points, instead of curvature, toreduce points initially. After that, we use uniform grid in the resampling process toretain feature points. Experiments show that this method preserves the feature of thedata while avoiding blank areas.Finally, registrationmethods for point cloud are investigated. The three-dimensionalcoordinate conversion method and the existing methods for point cloud registrationare introduced. Then we propose a method for initial registration based on principalcomponents analysis. At last, we introduce theIterative Closest Point (ICP) algorithm.All the methods have been verified using experiments.The method to input pointcloud data files in multiple formats and to display the modelusing OpenGL in theVC6.0environment are explained.The thesis is concluded with a summary and a prospect of the future work.