Outlier Detection of Scattered Point Cloud by Classification

Outlier detection and filtering are essential to reconstruct high quality surface in scattered point cloud preprocessing.In this paper a region growing algorithm for scattered point cloud and a new surface variation based local outlier factor(SVLOF) are proposed.They are employed to detect outliers from scattered point cloud.The reason for outlier's generation is analyzed and outliers are then classified to far outliers and near outliers based on their distances to the main body of the point cloud.Different algorithms are used to detect different kinds of outliers: for far outliers,region growing is used and for near outliers,SVLOF is used.The algorithm's ability to detect isolated and small group outliers is assessed by experiments based on emulation data and real scan data.