基于维度特征的RANSAC建筑物点云分割算法

提出一种以建筑物表面形状特征为分割依据的改进RANSAC点云分割算法。该算法以主成分分析算法为基础计算维度特征和熵函数,并以熵函数最小准则确定最优邻域,继而进行表面形状分类,运用法向量夹角作为约束条件对分类结果进行优化。将分类结果作为随机抽样一致性(RANSAC)点云分割算法的模型选择依据,进行建筑物表面分割,采用法向量和距离等约束条件对分割结果进行优化,从而分割出具有不同形状的特征表面。实验表明:文中提出的改进的RANSAC点云分割算法是可行的,能有效保留表面特征。

An improved RANSAC algorithm based on dimension feature in 3D point cloud of buildings

An improved RANSAC algorithm for complex surface shape feature in buildings is proposed in this paper . The dimension characteristics and the entropy functions are calculated , and the optimal neighborhood is determined by the criterion of the minimum value of entropy function ,and the surface shape is classified .T he precision of the classification result is improved by the angle of the normal vectors . Models are selected in random sampling (RANSAC) algorithm based on the classification results ,then the surfaces of building are segmented .The more precise surfaces are segmented by the normal vector and distance constraints .Experiment shows that the improved RANSAC algorithm is feasible ,and the surface characteristics can be effectively kept .