Research On Point Cloud Simplification Method For Feature Fitting

As the rapid development of data measurement technology and computer visualization technology, more and more methods of diversified data acquisition appeared. The introduction of 3D laser scanning technology makes the 3D reconstruction based on the mass point cloud data possible. Although it has brought a lot of convenience for the measure, the measure data volume increases sharply. Generally speaking, it can reach hundreds of thousands of points, or even millions of data points. The measurement data set is too large, and there are some redundant point cloud data in there which not only affect the subsequent data processing, storage, display and transmission, but also affect smoothness of object reconstructed surface. So it is very necessary to simplify the point cloud and to maintain effective information of the object being scanned. It helps to reduce the amount of data point cloud data processing, and to reduce the demand for computers and other hardware, in order to achieve the speed and efficiency of 3D modeling.In line with the requirement of precision and efficiency of surface reconstruction, this paper presents point cloud simplification algorithm based on the normal vector for feature fitting and based on feature points constraint for feature fitting. The main contents include the following aspects:(1)According to the principle of building spatial topological relationship of point cloud, this paper uses the rules different from the traditional k- d tree classifying. It chooses the longest axis as separate axis every time, and reduces the error of each axis direction through segmenting repeatedly.(2)This paper presents point cloud simplification algorithm based on the normal vector for feature fitting. First of all, establishing k- d tree of the point cloud data, by k neighborhood fitting out the plane to obtain the normal vector, then using vector as attributes to retain feature points, simplifying point cloud.(3)This paper presents point cloud simplification algorithm based on feature points constraint for feature fitting. It uses neighborhood curvature and the distance ratio as attributes, and then according to the distribution of feature points, sets a reasonable threshold retention of extraction of feature points, simplifies the rest of the feature points adaptive.Experimental results show that the point cloud simplification algorithm based on the normal vector oriented feature fitting and based on feature points constraint oriented feature fitting, can reduce the amount of data and ensure accuracy of subsequent point cloud, improve the efficiency of the characteristics of the fitting. But in the aspects of simplifying the proportion, retaining feature points and fitting effect of subsequent characteristics, point cloud simplification algorithm based on feature points constraint for feature fitting is superior to the point cloud simplification algorithm based on normal vector for feature fitting and the simplified algorithm of traditional bounding box.