Learning a Local Feature Descriptor for 3D LiDAR Scans

Dewan, Ayush; Caselitz, Tim; Burgard, Wolfram

doi:10.1109/iros.2018.8594420

Cited by 25 publications

(10 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In recent years, convolutional neural networks (CNNs) have become the state-of-the-art method for generating learning-based descriptors, owing to their ability to find complex patterns in data (Krizhevsky et al, 2012). For 3D point clouds, methods based on CNNs achieve impressive performance in applications such as object detection (Engelcke et al, 2017; Fang et al, 2015; Li et al, 2016; Maturana and Scherer, 2015; Qi et al, 2017; Riegler et al, 2017; Wohlhart and Lepetit, 2015; Wu et al, 2015), semantic segmentation (Graham et al, 2018; Li et al, 2016; Qi et al, 2017; Riegler et al, 2017; Tchapmi et al, 2017; Wu et al, 2018), and 3D object generation (Wu et al, 2016), and LiDAR-based local motion estimation (Dewan et al, 2018; Velas et al, 2018).…”

Section: Related Workmentioning

confidence: 99%

SegMap: Segment-based mapping and localization using data-driven descriptors

Dubé

Cramariuc

Dugas

et al. 2019

The International Journal of Robotics Research

166

View full text Add to dashboard Cite

Precisely estimating a robot’s pose in a prior, global map is a fundamental capability for mobile robotics, e.g., autonomous driving or exploration in disaster zones. This task, however, remains challenging in unstructured, dynamic environments, where local features are not discriminative enough and global scene descriptors only provide coarse information. We therefore present SegMap: a map representation solution for localization and mapping based on the extraction of segments in 3D point clouds. Working at the level of segments offers increased invariance to view-point and local structural changes, and facilitates real-time processing of large-scale 3D data. SegMap exploits a single compact data-driven descriptor for performing multiple tasks: global localization, 3D dense map reconstruction, and semantic information extraction. The performance of SegMap is evaluated in multiple urban driving and search and rescue experiments. We show that the learned SegMap descriptor has superior segment retrieval capabilities, compared with state-of-the-art handcrafted descriptors. As a consequence, we achieve a higher localization accuracy and a 6% increase in recall over state-of-the-art handcrafted descriptors. These segment-based localizations allow us to reduce the open-loop odometry drift by up to 50%. SegMap is open-source available along with easy to run demonstrations.

show abstract

Section: Related Workmentioning

confidence: 99%

SegMap: Segment-based mapping and localization using data-driven descriptors

Dubé

Cramariuc

Dugas

et al. 2019

The International Journal of Robotics Research

166

View full text Add to dashboard Cite

show abstract

“…For the lack of high repeatability issues, global descriptors, such as point feature histograms (PFH) [18] and viewpoint feature histograms (VFH) [19], have been proposed for using valuable techniques to extract features from point clouds. Recently, researchers tend to apply convolutional neural networks (CNN) to learn feature descriptors and to match their metrics in a uniform manner [20,21]. However, the limitation of using the deep learning method is that a large amount of training data is required, and when the similarity between the training data and the application environment is low, they cannot achieve good results.…”

Section: Related Workmentioning

confidence: 99%

Optimized LOAM Using Ground Plane Constraints and SegMatch-Based Loop Detection

Liu

Zhang

Qin

et al. 2019

Sensors

View full text Add to dashboard Cite

Reducing the cumulative error in the process of simultaneous localization and mapping (SLAM) has always been a hot issue. In this paper, in order to improve the localization and mapping accuracy of ground vehicles, we proposed a novel optimized lidar odometry and mapping method using ground plane constraints and SegMatch-based loop detection. We only used the lidar point cloud to estimate the pose between consecutive frames, without any other sensors, such as Global Positioning System (GPS) and Inertial Measurement Unit (IMU). Firstly, the ground plane constraints were used to reduce matching errors. Then, based on more accurate lidar odometry obtained from lidar odometry and mapping (LOAM), SegMatch completed segmentation matching and loop detection to optimize the global pose. The neighborhood search was also used to accomplish the loop detection task in case of failure. Finally, the proposed method was evaluated and compared with the existing 3D lidar SLAM methods. Experiment results showed that the proposed method could realize low drift localization and dense 3D point cloud map construction.

show abstract

“…Local 3D feature descriptors are evaluated in [2] to merge LiDAR maps from a ground vehicle to dense maps from monocular SLAM captured by an aerial vehicle acknowledging the robustness of SHOT [5]. Learned 3D descriptors and matching metrics [21]- [23] show promising results although require powerful hardware, which limits the implementation on resource constrained vehicles. SHOT enriched with LiDAR intensity data is used in [24] along with a probabilistic selection scheme for multi-session localization.…”

Section: Related Workmentioning

confidence: 99%

Relocalization With Submaps: Multi-Session Mapping for Planetary Rovers Equipped With Stereo Cameras

Giubilato

Vayugundla

Schuster

et al. 2020

IEEE Robot. Autom. Lett.

View full text Add to dashboard Cite

To enable long term exploration of extreme environments such as planetary surfaces, heterogeneous robotic teams need the ability to localize themselves on previously built maps. While the Localization and Mapping problem for single sessions can be efficiently solved with many state of the art solutions, place recognition in natural environments still poses great challenges for the perception system of a robotic agent. In this paper we propose a relocalization pipeline which exploits both 3D and visual information from stereo cameras to detect matches across local point clouds of multiple SLAM sessions. Our solution is based on a Bag of Binary Words scheme where binarized SHOT descriptors are enriched with visual cues to recall in a fast and efficient way previously visited places. The proposed relocalization scheme is validated on challenging datasets captured using a planetary rover prototype on Mount Etna, designated as a Moon analogue environment.

show abstract

Learning a Local Feature Descriptor for 3D LiDAR Scans

Cited by 25 publications

References 19 publications

SegMap: Segment-based mapping and localization using data-driven descriptors

SegMap: Segment-based mapping and localization using data-driven descriptors

Optimized LOAM Using Ground Plane Constraints and SegMatch-Based Loop Detection

Relocalization With Submaps: Multi-Session Mapping for Planetary Rovers Equipped With Stereo Cameras

Contact Info

Product

Resources

About