面向虚拟测方系统的点云聚类与拟合理论

发布时间：2018-11-18 11:03

【摘要】：针对现有点云聚类方法及曲面拟合方法的不足:（1）聚类中心及特征的选择导致的曲面片过分割问题，蚁群优化参数的改变对聚类有效性的影响问题；（2）径向基函数中法向离面约束点对曲面插值误差的影响问题。研究提出了改进的蚁群优化投影寻踪聚类方法和基于点云密度的径向基函数曲面插值法。论文首先构建了集点云采集、无线传输和入库为一体的虚拟测方系统，，在分析误差来源与坐标系统关系的基础上推导出点云坐标统一的误差传播模型，为后续的分析提供数据源。然后提出了改进的蚁群算法优化投影寻踪法来实现点云聚类，构建了蚁群算法中信息素系数更新的反正切函数模型，采用DUNN指数法定量评价了蚁群优化不同聚类算法有效性，研究表明优化聚类改善了曲面过分割结果。最后针对优化聚类后的点云对象，提出基于K邻域点云密度的径向基函数曲面插值方法，挖掘了曲面插值点个数与法向离面约束点对基于点云密度的径向基曲面插值精度的定量影响，分析了二次曲面拟合、Nurbs曲面插值与基于点云密度的径向基函数插值精度对体积的误差影响，研究表明本文方法具有较高的曲面拟合精度。
[Abstract]:The shortcomings of existing point cloud clustering methods and surface fitting methods are as follows: (1) the problem of over-segmentation caused by the selection of clustering centers and features and the effect of ant colony optimization parameters on clustering effectiveness; (2) the influence of normal off-plane constraint points on interpolation error in radial basis function. An improved projection pursuit clustering method for ant colony optimization and a radial basis function surface interpolation method based on point cloud density are proposed. Firstly, a virtual square measurement system is constructed, which integrates point cloud collection, wireless transmission and storage. Based on the analysis of the relationship between the error source and coordinate system, a unified error propagation model of point cloud coordinates is derived. Provide data sources for subsequent analysis. Then the improved ant colony optimization projection pursuit method is proposed to realize point cloud clustering, and an inverse tangent function model of pheromone coefficient updating in ant colony algorithm is constructed. The validity of different ant colony optimization clustering algorithms is evaluated quantitatively by DUNN exponent method. The results show that the optimal clustering improves the results of surface hypersegmentation. Finally, a radial basis function (RBF) surface interpolation method based on K-neighborhood point cloud density is proposed for the optimized point cloud objects. The quantitative effects of the number of surface interpolation points and the normal off-plane constraint points on the interpolation accuracy of radial basis surface based on point cloud density are excavated, and the Quadric surface fitting is analyzed. The effect of the accuracy of Nurbs surface interpolation and radial basis function interpolation based on point cloud density on the volume error is studied. The results show that the proposed method has high surface fitting accuracy.
【学位授予单位】：中国矿业大学（北京）
【学位级别】：博士
【学位授予年份】：2014
【分类号】：P23

【参考文献】