%0 Conference Paper %1 IAS2022 %A Neumann, M. %A Borrmann, D. %A Nüchter, A. %B Proceedings of the 17th Conference on Intelligent Autonomous Systems (IAS '22) %C Zagreb, Croatia %D 2023 %K imported %N 577 %P 342--359 %R 10.1007/978-3-031-22216-0_24 %T Semantic Classification in Uncolored 3D Point Clouds using Multiscale Features %U https://robotik.informatik.uni-wuerzburg.de/telematics/download/ias2022.pdf %X While the semantic segmentation of 2D images is already a well-researched field, the assignment of semantic labels to 3D data is lagging behind. This is partly due to the fact that prelabeled training data is only rarely available since not only the training and application of classification methods but also the manual labeling process are much more time-consuming in 3D. This paper focuses on the more classical approach of first calculating features and subsequently applying a classification algorithm. Existing handcrafted feature definitions are enhanced by using multiple selected reductions of the point cloud as approximations. This serves as input to train a well-studied random forest classifier. A comparison to a recently presented deep learning approach, i.e., the Kernel Point Convolution method, reveals that there are well-justified applications for both modern and classical machine learning methods. To enable the smooth conversion of existing 3D scenes to semantically labeled 3D point clouds the tool Blender2Helios is presented. We show that the therewith generated artificial data is a good choice for training real-world classifiers.

@inproceedings{IAS2022, abstract = {While the semantic segmentation of 2D images is already a well-researched field, the assignment of semantic labels to 3D data is lagging behind. This is partly due to the fact that prelabeled training data is only rarely available since not only the training and application of classification methods but also the manual labeling process are much more time-consuming in 3D. This paper focuses on the more classical approach of first calculating features and subsequently applying a classification algorithm. Existing handcrafted feature definitions are enhanced by using multiple selected reductions of the point cloud as approximations. This serves as input to train a well-studied random forest classifier. A comparison to a recently presented deep learning approach, i.e., the Kernel Point Convolution method, reveals that there are well-justified applications for both modern and classical machine learning methods. To enable the smooth conversion of existing 3D scenes to semantically labeled 3D point clouds the tool \emph{Blender2Helios} is presented. We show that the therewith generated artificial data is a good choice for training real-world classifiers.}, added-at = {2022-05-31T08:24:24.000+0200}, address = {Zagreb, Croatia}, author = {Neumann, M. and Borrmann, D. and N{\"u}chter, A.}, biburl = {https://www.bibsonomy.org/bibtex/290645003c56e0c20862560971932dad1/nuechter76}, booktitle = {Proceedings of the 17th Conference on Intelligent Autonomous Systems (IAS '22)}, doi = {10.1007/978-3-031-22216-0_24}, interhash = {f4fae22b9fb5e3e5047ca71ca83d7a38}, intrahash = {90645003c56e0c20862560971932dad1}, keywords = {imported}, month = {June}, number = 577, optnumber = {192}, optpages = {173--182}, optpublisher = {Springer}, optseries = {Advances in Intelligent Systems and Computing}, optvolume = {2}, pages = {342--359}, series = {Lecture Notes in Networks and Systems}, timestamp = {2023-01-29T16:42:49.000+0100}, title = {{Semantic Classification in Uncolored 3D Point Clouds using Multiscale Features}}, url = {https://robotik.informatik.uni-wuerzburg.de/telematics/download/ias2022.pdf}, year = 2023 }