基于标记的多状态离群点去除算法

Marker-based multi-state outlier removal algorithm

针对点云采数据集过程中由于干扰因素产生的离群点及离群点簇,提出一种基于标记的多状态离群点去除方法。首先,通过基于正交分量比值方法标记孤立离群点;其次,运用改进的DBSCAN聚类方法对已标记的点云数据进行聚类;然后,统计各个聚类中的已标记离群点占该聚类中点的数量比例,将大于设定阈值的聚类视为离群点簇,将其和剩余标记的孤立离群点进行删除。实验表明,该方法不仅能够去除原始点云数据中孤立的离群点,而且可以有效去除空间中成簇的离群点,为后续点云处理奠定了有利的基础。

Aiming at outliers and outlier clusters generated by disturbance factors during points cloud data collection,this paper proposes a marker based multi-state outlier removal method.Firstly,isolated outliers are marked by orthogonal component ratio method.Secondly,the improved DBSCAN clustering method is used to cluster the marked points cloud data.Then,the percentage of marked outliers in each cluster in the number of points in the cluster is counted,and the cluster larger than the set threshold is regarded as an outlier cluster,which is deleted together with the remaining marked outliers.Experiments show that this method can not only remove the isolated outliers in the original points cloud data,but also effectively remove the cluster outliers in the space,which lays a favorable foundation for subsequent point cloud processing.