Research And Applications On Fitting Multiple Curves To Point Cloud With Complicated Topological Structures

Point cloud data is a kind of discrete data which shows surface of object. It is easy to obtain and has rich details. But the discrete feature makes it not easy to use point cloud data in reverse engineering. Research on point cloud curve fitting is one of the most popular research subjects.Common algorithms for point cloud curve fitting can’t handle complex topological structure. By studying the background and the development of the point cloud curve fitting, we analyze domestic and foreign research on this subject. On the basis of the research before we put forward a point cloud classification and curve fitting algorithm based on PCA. The algorithm bases on splitting-merging thought. Firstly, the point cloud is divided into multiple PCA units, and then based on some conditions set before hand, these units are merged to get the final classification results. On the basis of classification results, B-spline curves are fitting. SDM algorithm is used to fit the curves. This algorithm can fit curves on point cloud and meanwhile makes a classification. This method works on point cloud which has complicated topological structures.Point cloud curve fitting results can be used as a skeleton. A skeleton driven fast point cloud deformation algorithm is put forward. The idea of this algorithm is binding the point cloud data with the skeleton points, interactively deforming the skeleton points based on quadratic B-spline curve, and then deforming the point cloud data binding with these skeleton points. The steps of this algorithm is clear, easy to implement and fast. Based on the classification and curve fitting method we put forward, some study on skeleton extraction of characters is shown. This paper proposes a Chinese characters skeleton extraction algorithm based on point cloud. First we convert Chinese characters to point cloud, and then we use the classification and curve fitting algorithm to extract the skeleton of Chinese characters. This algorithm can reduce the effect of noise on characters. And skeleton extracted is smooth and can reflect the topological characteristics of Chinese characters.Based on the research on classification and curve fitting of planar point cloud data, point cloud deformation based on skeleton and the study of Chinese character feature extraction, we propose three algorithms. The classification and curve fitting method has obvious advantages on point cloud with complex topology. Algorithms of rapid point cloud deformation based on skeleton and feature extraction of Chinese characters based on point cloud, have some inspiration for future research.