The TUBS Road User Dataset: A New LiDAR Dataset and its Application to CNN-based Road User Classification for Automated Vehicles

Plachetka, Christopher; Rieken, Jens; Maurer, Markus

doi:10.1109/itsc.2018.8569765

Cited by 10 publications

(5 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The international standards for 3D modeling are set by the OGC CityGML model [14], which defines the principles for mapping physical topographic objects and physical volumetric objects [14,18]. Physical volumetric models are combined with semantic and functional data to develop detailed models of the urban systems [18,22], road networks and other systems, including in real time [23,24].…”

Section: Literature Reviewmentioning

confidence: 99%

Three-Dimensional Modeling and Visualization of Single Tree LiDAR Point Cloud Using Matrixial Form

Tarsha Kurdi,

Lewandowicz,

Shan

et al. 2024

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

Tree modeling and visualization still represent a challenge in the Light Detecting And Ranging (LiDAR) area. Starting from the segmented tree point clouds, this paper presents an innovative tree modeling and visualization approach. The algorithm simulates the tree point cloud by a rotating surface. Three matrixes, X, Y, and Z are calculated by considering the middle of the projected tree point cloud on the horizontal plane. This mathematical form not only allows tree modeling and visualization, but also permits the calculation of geometric characteristics and parameters of the tree. The superimposition of the tree point cloud over the constructed model confirms its high accuracy where all the points of the tree cloud are within the constructed model. The tests with multiple single trees demonstrate an overall average fit between 0.3 and 0.89 m. The built tree models are also compliant with the Open Geospatial Consortium (OCG) City GML standards at the level of a physical model. This approach opens a door to numerous applications for visualization, computation, and study of forestry and vegetation in urban as well as rural areas.

show abstract

Section: Literature Reviewmentioning

confidence: 99%

Three-Dimensional Modeling and Visualization of Single Tree LiDAR Point Cloud Using Matrixial Form

Tarsha Kurdi,

Lewandowicz,

Shan

et al. 2024

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

show abstract

“…(Dewan et al, 2016) use a Bayesian approach and motion cues to automatically detect and track objects. (Plachetka et al, 2018) track manually annotated 3D bounding boxes and corresponding polygons (for RGB segmentation). (Chang et al, 2019) accumulate manually selected points over time to automatically infer fixed size 3D bounding boxes.…”

Section: Related Workmentioning

confidence: 99%

“…Algorithmically guided generation of pixel-level instance segmentation masks in both 2D and 3D, with 2D bounding boxes as a by-product, speeding up annotation time by a factor of up to 8.55. (Marion et al, 2018) mesh + reconstruction 6DoF RGB, D ( ) (Suchi et al, 2019) incremental scene building 2D D -- (Wong et al, 2015) learning 4DoF RGB - (Monica et al, 2017) graph segmentation -D - (Zimmer et al, 2019) interpolation 4DoF -- (Wang et al, 2019) tracking + clustering 4DoF -- (Arief et al, 2020) tracking + clustering 4DoF -- (Plachetka et al, 2018) tracking + optimization 4DoF, 2D RGB, D (Lee et al, 2018) learning 4DoF D - (Huang et al, 2020) 3D-2D projection -RGB, D - (Yan et al, 2020) clustering 3DoF -- (Hodaň et al, 2017) mesh + reconstruction 6DoF, 2D - (Xiang et al, 2018) mesh + reference frame 6DoF D (Grenzdörffer et al, 2020) mesh + reference frame 6DoF, 2D D (Xie et al, 2016) label transfer -RGB (Patil et al, 2019) interpolation 4DoF - (Dewan et al, 2016) Bayesian approach 3DoF D (Chang et al, 2019) accumulation over time 6DoF -…”

Section: Introductionmentioning

confidence: 99%

SALT: A Semi-automatic Labeling Tool for RGB-D Video Sequences

Stumpf,

Krauß,

Reis

et al. 2021

Preprint

View full text Add to dashboard Cite

Large labeled data sets are one of the essential basics of modern deep learning techniques. Therefore, there is an increasing need for tools that allow to label large amounts of data as intuitively as possible. In this paper, we introduce SALT, a tool to semi-automatically annotate RGB-D video sequences to generate 3D bounding boxes for full six Degrees of Freedom (DoF) object poses, as well as pixel-level instance segmentation masks for both RGB and depth. Besides bounding box propagation through various interpolation techniques, as well as algorithmically guided instance segmentation, our pipeline also provides built-in pre-processing functionalities to facilitate the data set creation process. By making full use of SALT, annotation time can be reduced by a factor of up to 33.95 for bounding box creation and 8.55 for RGB segmentation without compromising the quality of the automatically generated ground truth.

show abstract

“…• Usage of an adaptive ground surface estimation, which serves as the basis for a surface-relative environment representation, • a multi-stage pointcloud preprocessing pipeline, including pointcloud compression and clustering, • a motion classification of detected clusters, and • a hybrid representation for movable and stationary environment elements, consisting of a multi-layer semantic grid map and an IMM-EKF-based object tracking with explicit modeling of partial occlusions. We present the latest additions to our previous work (e. g. [3]- [10]) and provide evaluation results of the perception performance using a subset of the TUBS Road User Dataset [11]. We extend common evaluation strategies by movable elements stationary elements traversable regions Figure 1.…”

Section: A Contribution Of This Papermentioning

confidence: 99%

“…Despite the fact that using such a publicly available dataset would have allowed the comparison with other approaches, both expose some drawbacks regarding their usability for evaluation of the presented perception system (i. e., unlabeled regions or ranges [36], no per-scan annotated data [75]). Thus, we use a preliminary version of the TUBS Road User Dataset [11] for evaluating our approach. This dataset contains labeled objects within the entire field of view of the sensor and for each LiDAR scan, recorded along the project route of the project Stadtpilot.…”

Section: Definition Of Datasets and Evaluation Criteriamentioning

confidence: 99%

A LiDAR-based real-time capable 3D Perception System for Automated Driving in Urban Domains

Rieken,

Maurer

2020

Preprint

Self Cite

View full text Add to dashboard Cite

We present a LiDAR-based and real-time capable 3D perception system for automated driving in urban domains. The hierarchical system design is able to model stationary and movable parts of the environment simultaneously and under real-time conditions. Our approach extends the state of the art by innovative in-detail enhancements for perceiving road users and drivable corridors even in case of nonflat ground surfaces and overhanging or protruding elements. We describe a runtime-efficient pointcloud processing pipeline, consisting of adaptive ground surface estimation, 3D clustering and motion classification stages.Based on the pipeline's output, the stationary environment is represented in a multi-feature mapping and fusion approach. Movable elements are represented in an object tracking system capable of using multiple reference points to account for viewpoint changes. We further enhance the tracking system by explicit consideration of occlusion and ambiguity cases.Our system is evaluated using a subset of the TUBS Road User Dataset. We enhance common performance metrics by considering application-driven aspects of real-world traffic scenarios. The perception system shows impressive results and is able to cope with the addressed scenarios while still preserving real-time capability.

show abstract

The TUBS Road User Dataset: A New LiDAR Dataset and its Application to CNN-based Road User Classification for Automated Vehicles

Cited by 10 publications

References 15 publications

Three-Dimensional Modeling and Visualization of Single Tree LiDAR Point Cloud Using Matrixial Form

Three-Dimensional Modeling and Visualization of Single Tree LiDAR Point Cloud Using Matrixial Form

SALT: A Semi-automatic Labeling Tool for RGB-D Video Sequences

A LiDAR-based real-time capable 3D Perception System for Automated Driving in Urban Domains

Contact Info

Product

Resources

About