The Research On Registration Algorithm Of 3D Data

With the development of laser measurement technology, more choices for 3D surface digitalizing are provided so that registration of data technology, which are from different measurement equipments and views, is playing more and more important role. In developing digital remanufacturing and repairing fields, it is an important technology part to compare the data between designed model and measured model. Therefore, registration technology is studied and some results are obtained as follows:Firstly, considering the noise data in the original scanned data, octree data structure is used to remove the noise data and ensure the accuracy of following works. Meanwhile, in order to simplify and improve the efficiency of registration, feature information of surface is utilized. As the original position of point cloud may cause the unconvergence of registration, a special rotation based on primary component analysis is calculated to regulate the original position. Secondly, the key way to improve ICP algorithm is to advance the efficiency of searching the corresponding points in different data sets. Based on Hash geometric, Gaussian sphere registering algorithms, genetic algorithm and iterative approach, a novel registration algorithm, which gravity is calculated in sections, is presented so that the efficiency of searching the corresponding points and accuracy are improved. For Genetic Algorithm has a good perform on probability search and optimization, it is employed to gain a better initial rotation matrix for iterative registration.Finally, to provide data base for robot path planning, the presented registration scheme is applied to a digital repairing robots system, the surface damage of a work piece is separated by registering the scanned data with CAD model. The result shows that the presented scheme is successful.