TU Darmstadt / ULB / TUbiblio

Unsupervised 4D LiDAR Moving Object Segmentation in Stationary Settings with Multivariate Occupancy Time Series

Kreutz, Thomas ; Mühlhäuser, Max ; Sanchez Guinea, Alejandro (2023)
Unsupervised 4D LiDAR Moving Object Segmentation in Stationary Settings with Multivariate Occupancy Time Series.
2023 IEEE/CVF Winter Conference on Applications of Computer Vision. Hawaii, USA (03.01.2023-07.01.2023)
doi: 10.1109/WACV56688.2023.00169
Konferenzveröffentlichung, Bibliographie

Kurzbeschreibung (Abstract)

In this work, we address the problem of unsupervised moving object segmentation (MOS) in 4D LiDAR data recorded from a stationary sensor, where no ground truth annotations are involved. Deep learning-based state-of- the-art methods for LiDAR MOS strongly depend on anno- tated ground truth data, which is expensive to obtain and scarce in existence. To close this gap in the stationary set- ting, we propose a novel 4D LiDAR representation based on multivariate time series that relaxes the problem of un- supervised MOS to a time series clustering problem. More specifically, we propose modeling the change in occupancy of a voxel by a multivariate occupancy time series (MOTS), which captures spatio-temporal occupancy changes on the voxel level and its surrounding neighborhood. To perform unsupervised MOS, we train a neural network in a self- supervised manner to encode MOTS into voxel-level feature representations, which can be partitioned by a clustering al- gorithm into moving or stationary. Experiments on station- ary scenes from the Raw KITTI dataset show that our fully unsupervised approach achieves performance that is com- parable to that of supervised state-of-the-art approaches.

Typ des Eintrags: Konferenzveröffentlichung
Erschienen: 2023
Autor(en): Kreutz, Thomas ; Mühlhäuser, Max ; Sanchez Guinea, Alejandro
Art des Eintrags: Bibliographie
Titel: Unsupervised 4D LiDAR Moving Object Segmentation in Stationary Settings with Multivariate Occupancy Time Series
Sprache: Englisch
Publikationsjahr: 22 Januar 2023
Verlag: IEEE
Buchtitel: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV'23)
Veranstaltungstitel: 2023 IEEE/CVF Winter Conference on Applications of Computer Vision
Veranstaltungsort: Hawaii, USA
Veranstaltungsdatum: 03.01.2023-07.01.2023
DOI: 10.1109/WACV56688.2023.00169
Zugehörige Links:
Kurzbeschreibung (Abstract):

In this work, we address the problem of unsupervised moving object segmentation (MOS) in 4D LiDAR data recorded from a stationary sensor, where no ground truth annotations are involved. Deep learning-based state-of- the-art methods for LiDAR MOS strongly depend on anno- tated ground truth data, which is expensive to obtain and scarce in existence. To close this gap in the stationary set- ting, we propose a novel 4D LiDAR representation based on multivariate time series that relaxes the problem of un- supervised MOS to a time series clustering problem. More specifically, we propose modeling the change in occupancy of a voxel by a multivariate occupancy time series (MOTS), which captures spatio-temporal occupancy changes on the voxel level and its surrounding neighborhood. To perform unsupervised MOS, we train a neural network in a self- supervised manner to encode MOTS into voxel-level feature representations, which can be partitioned by a clustering al- gorithm into moving or stationary. Experiments on station- ary scenes from the Raw KITTI dataset show that our fully unsupervised approach achieves performance that is com- parable to that of supervised state-of-the-art approaches.

Freie Schlagworte: emergenCITY, emergenCITY_INF
Fachbereich(e)/-gebiet(e): 20 Fachbereich Informatik
20 Fachbereich Informatik > Telekooperation
LOEWE
LOEWE > LOEWE-Zentren
LOEWE > LOEWE-Zentren > emergenCITY
Hinterlegungsdatum: 23 Jan 2023 13:04
Letzte Änderung: 19 Jan 2024 19:30
PPN:
Export:
Suche nach Titel in: TUfind oder in Google
Frage zum Eintrag Frage zum Eintrag

Optionen (nur für Redakteure)
Redaktionelle Details anzeigen Redaktionelle Details anzeigen