Print Email Facebook Twitter A voxel-based methodology to detect (clustered) outliers in aerial lidar point clouds Title A voxel-based methodology to detect (clustered) outliers in aerial lidar point clouds Author Griffioen, Simon (TU Delft Architecture and the Built Environment) Contributor Peters, Ravi (mentor) Ledoux, Hugo (mentor) Meijers, Martijn (graduation committee) Degree granting institution Delft University of Technology Programme Geomatics Date 2018-11-12 Abstract To obtain 3D information of the Earth’s surface, airborne LiDAR technologyis used to quickly capture high-precision measurements of the terrain.Unfortunately, laser scanning techniques are prone to producing outliersand noise (i.e. wrong measurements). Therefore, a pre-process of the pointcloud is required to detect and remove spurious measurements. While outlierdetection in datasets has been extensively researched, in 3D point clouddata it is still an ongoing problem. Especially, clustered outliers are hard todetect with previous local-neighborhood based algorithms.This research explores the possibilities of using a voxel-based approach toautomatically remove outliers from aerial point clouds. A workflow is designedin which a series of voxel-based operations are integrated, with theaim to detect all types of outliers and minimize false positives. Voxels canbe processed more efficiently than 3D points for two reasons: (1) A voxelgridcan be analyzed using efficient image processing techniques; (2) Voxelsgroup inner points before feature extraction using neighborhood operators.Outliers are detected in two steps. First, the source point cloud is voxelized.Secondly, outliers are detected by computing connected components and labelingvoxels not connected to the largest region as outliers. Simultaneously,analysis of the point’s local density, shape (planar) and intensity minimizeclassification of false positives.The presented algorithm generally detects outliers with a higher accuracythan previous local neighborhood-based methods. A comparison with anexisting approach shows that more outliers are detected. Above all, clusteredoutliers are removed. However, some issues can still be improved.First, more research is necessary to classify outliers based on non-arbitrarydecisions. This could potentially be improved by introducing supervisedlearning algorithms. Secondly, more attention is required to process massivepoint clouds that do not fit in internal memory. This study proposes apossible streaming solution. Subject point cloudaerial lidarOutlier Detectionvoxelconnected components labeling To reference this document use: http://resolver.tudelft.nl/uuid:2ffa73f4-34cc-4ea0-82df-11e61cb47bea Part of collection Student theses Document type master thesis Rights © 2018 Simon Griffioen Files PDF P5_4131940.pdf 11.12 MB PDF P5_4131940_presentation.pdf 8.07 MB PDF P2_4131940.pdf 974.28 KB Close viewer /islandora/object/uuid:2ffa73f4-34cc-4ea0-82df-11e61cb47bea/datastream/OBJ2/view