结合上下文特征和图割算法的车载点云聚类方法

可靠、准确的点云聚类是后续高精度场景目标分析与解译的基础.该文提出了一种基于上下文特征和图割算法的车载点云聚类方法.首先用DBSCAN （density-based spatial clustering of applications with noise）对点云数据进行过分割，得到密度可达的超体素；然后引入空间和属性上下文特征来描述超体素间的关联，并用于定义超体素构建的图模型边的权值；最后基于多标记的图割优化算法得到最佳超体素聚簇.实验结果表明，该方法能够有效改善点云聚类过分割，从而提高聚类的精度.

Vehicle Point Cloud Clustering Based on Contextual Feature and Graph Cut

Reliable and accurate clustering of point cloud is the basis for subsequent high-precision analysis and interpretation of scene target. This paper presents a method for vehicle point cloud segment using supervoxel and graph-cut with contextual feature. Firstly, density-based spatial clustering of applications with noise (DBSCAN) is used to segment the point cloud data, and density-reachable supervoxels can be obtained. Secondly, spatial and attribute contextual features are introduced to describe the correlation between supervoxels and to define the weights of the edges of the graph model constructed by supervoxels. Finally, the optimal supervoxel clustering is obtained based on multi-label graph-cut optimization algorithm. Experimental results show that the proposed method has improved accuracy and performance on over segmentation in clustering.