Segmentation of conifer tree crowns from terrestrial laser scanning point clouds in mixed stands of Scots pine and Norway spruce

Abstract Terrestrial laser scanning of conifer tree crowns is challenged by occlusion problems causing sparse point clouds for many trees. Automatic segmentation from a task that has only recently been addressed and not solved in way all trees can be segmented automatically without assignment errors. We developed new algorithm based on region growing seeds voxelized 3D clouds. In our data, field measured positions diameters were available as input data to estimate crown cores the growing. other applications, these derived cloud. Segmentation success was judged visually voxel 1294 Norway spruce Scots pine 24 plots six mixed species stands. Only about half had minor or no errors allowing fit concentric models. most often caused unsegmented neighbors at edge sample plots. Wrong assignments parts also more frequent dense groups understory For some trees, too describe crown. rates considerably higher dominant plot center. Despite incomplete automatic crowns, metrics describing size shape could large number A description irregular possible due upper