摘要
为了在保持特征的基础上有效地简化点云数据,提出了基于聚类的点云精简算法.对点云进行三维栅格剖分,在每个栅格中选取1个代表点作为初始类核心,然后将点云中其他数据点归入欧氏距离最近的初始类中,遍历各个类,若类内某两点的法向量偏差大于给定带宽则对该类进行迭代细分,并对各个类进行均值漂移处理,将得到的局部模态点取代该类,从而实现点云简化.以手机外壳、人头、麻花钻为典型实例,对具有不同表面特征的点云数据进行了验证.结果表明,该算法能对点云数据进行直接而有效的精简,在曲率变化大、附加特征多的表面仍能很好地保留原始模型的几何形状.
To simplify the point cloud while preserving small features, a novel algorithm based on clustering is proposed. The whole point could is divided into a series of initial sub-clusters with the 3-D grid subdivision method, and in each initial sub-cluster one representative point is selected as the centroid. The points other than those representatives are distributed to their nearest ini- tial sub-cluster centroids according to Euclidean distance, and new clusters are generated. Traversing all new formed clusters, if the normal vector deviation of any two inner points is greater than the given threshold, the cluster is necessarily subdivided, then each cluster is processed iteratively by mean shift to obtain the local mode points, which are adopted to substitute the clusters. Some typical cases with various surface features, such as mobile shell, human-head sculpture and twisted drill, are chosen to verify this method. The result indicates that the new algorithm enables to reduce data directly and efficiently while maintaining the geometry of the original model, especially for the surfaces with sharp edges and complex additional features.
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2010年第11期37-40,共4页
Journal of Xi'an Jiaotong University
基金
国家自然科学基金资助项目(50975219)
江苏省科技支撑计划资助项目(BE2008058)
关键词
点云简化
聚类
均值漂移
point cloud simplification
clustering
mean shift
作者简介
史宝全（1982-），男，博士生；梁晋（联系人），男，副教授．
