基于距离统计的有序纹理点云离群点检测

Outlier Detection Based on Distance Statistics for Ordered Texture Point Cloud

三维数据的离群点检测是纹理点云数据处理的重要内容之一,为了有效快速地检测离群点,根据纹理点云的有序结构特征,提出了基于距离统计的检测算法。首先在每个点到其K邻域中其他点距离的基础上计算出K邻域距离;然后根据有序点云中该距离符合正态分布的特点和正态分布3σ定理,将超出3倍方差范围的点认定为离群点。实验结果显示算法采用曼哈顿-最大距离进行检测,当K为4时可以更加快速准确地将有序点云中的离群点检测出来。由此得出,基于距离统计的算法可以有效地将离群点检测出来,同时成功地应用于纹理点云的离群点检测。

3D outlier detection is an important processing of texture point cloud,in order to effectively detect the outlier quickly, a outlier detection method based on distance statistics is proposed,according to the ordered structure characteristic of texture point cloud. K neighborhood distance of every point is calculated by the distances between the point and its every K neighborhood point firstly;and then as the K neighborhood distance of ordered point cloud follow the normal distribution and the normal distribution 3σ theorem,the point will be detected as outlier point if its K neighborhood distance is beyond 3σrange. The result of experiments show that the proposed method can more quickly and accurately to detect outlier,if Manhattan-Maximum distance is adapted and K is 4. The conclusion is that the outlier detection method based on distance statistics can effectively detect outliers,and is applied on texture point cloud successfully.