Research On Data Pre-Processing Technology Of Scattered Point Cloud Based On Curvatures Feature

With the continuous improvement of accuracy of the non-contact measurement equipment, non-contact measurement is increasingly used in aviation, aerospace, shipbuilding, biotechnology, equipment manufacturing and other engineering fields. Owing to the accuracy of scattered point cloud data acquired by non-contact measurement is vulnerable to the impact of the environment, a lot of error points exist in the point cloud, meanwhile, because of the high accuracy of equipment, causes there are a lot of redundant data in the point cloud. The existence of the error points will seriously affect the quality of the subsequent surface reconstruction, while a large number of redundant data makes the whole reconstruction work requires a great deal of computing resources, low efficiency, poor accuracy, therefore, the smoothing and reduction of scattered point cloud data are indispensable means of data pre-processing to ensure the quality of surface reconstruction.Based on the scattered point cloud data obtained by high-precision scanning, the technology of data smoothing and reduction which can retain the surface geometrical characteristics (the normal vector, curvature) data efficiently is studied in detail, moreover, the algorithms of data smoothing and reduction are programmed, and scattered point cloud with different scale and requirements are processed in this program to validate the rationality and effectiveness of these algorithms. Finally, a software with complete the algorithms of data smoothing and reduction is developed.The major research work of this thesis is as follows:1. Based on space partition technology, the normal k neighborhood search algorithm with whole expansion is improved by the searching technology of automatic one side expansion. Using the technology, the k neighboring points can be searched more quickly.2. The computing methods of geometric features of scattered point cloud, such as normal vector and curvature, are researched, including:the estimation and correction algorithm of normal vector, the estimation algorithm of curvature. These algorithms will be used in the subsequent data smoothing and streamlined process.3. In order to avoid the important boundary points lost, the extraction algorithm of boundary points from the scattered point cloud is introduced into the smoothing algorithm. Then, using the average curvature as a threshold value to partition the scattered points into smooth areas and saltation areas, and adaptive smoothing algorithms are used to these areas. The full set of smoothing algorithm can not only retain the characteristic information of the boundary geometry and information of curvature change, but also arranges scattered points more regular, orderly. These works lay a good foundation for the future surface reconstruction.4. The technologies of calculating correcting factor of the average curvature and adjusting data partition dynamically are studied in order to improve data reduction efficiency and retain the original geometric characteristic information properly. At the same time, the space-based segmentation curvature-based data reduction methods are integrated and used to smooth areas and saltation areas respectively. By using the data reduction algorithm to all kinds of scattered point cloud, the effectiveness of algorithm is validated.5. The all study results, such as the algorithms of K-neighborhood search, estimation and correction of the normal vector, estimation of curvature, extraction of the boundary points, data smoothing and data reduction are all integrated. With the use of the secondary development tool CAA of CATIA software, a program of data pre-processing is developed.