A Robust Gaussian Process-Based LiDAR Ground Segmentation Algorithm for Autonomous Driving

Jin, Xianjian; Yang, Hang; Liao, Xiaofei; Yan, Zeyuan; Wang, Qikang; Li, Zhiwei; Wang, Zhaoran

doi:10.3390/machines10070507

Cited by 4 publications

(5 citation statements)

References 28 publications

(44 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This paper takes point cloud clustering, bounding box fitting and point cloud classification based on deep learning as the technical route and proposes a vehicle detection algorithm based on the combination of clustering and deep learning, the detailed technical route is shown in Figure 1, in which we used the previous work [34] to obtain the drivable area. In simple terms, the ground segmentation algorithm based on the Gaussian process is used to remove the ground point cloud.…”

Section: Main Workmentioning

confidence: 99%

Robust LiDAR-Based Vehicle Detection for On-Road Autonomous Driving

Jin

Yang

et al. 2023

Remote Sensing

Self Cite

View full text Add to dashboard Cite

The stable detection and tracking of high-speed vehicles on the road by using LiDAR can input accurate information for the decision-making module and improve the driving safety of smart cars. This paper proposed a novel LiDAR-based robust vehicle detection method including three parts: point cloud clustering, bounding box fitting and point cloud recognition. Firstly, aiming at the problem of clustering quality degradation caused by the uneven distribution of LiDAR point clouds and the difference in clustering radius between point cloud clusters in traditional DBSCAN (TDBSCAN) obstacle clustering algorithms, an improved DBSCAN algorithm based on distance-adaptive clustering radius (ADBSCAN) is designed, and a point cloud KD-Tree data structure is constructed to speed up the traversal of the algorithm; meanwhile, the OPTICS algorithm is introduced to enhance the performance of the proposed algorithm. Then, by adopting different fitting strategies for vehicle contour points in various states, the adaptability of the bounding box fitting algorithm is improved; Moreover, in view of the shortcomings of the poor robustness of the L-shape algorithm, the principal component analysis method (PCA) is introduced to obtain stable bounding box fitting results. Finally, considering the time-consuming and low-accuracy training of traditional machine learning algorithms, advanced PointNet in deep learning technique is built to send the point cloud within the bounding box of a high-confidence vehicle into PointNet to complete vehicle recognition. Experiments based on our autonomous driving perception platform and the KITTI dataset prove that the proposed method can stably complete vehicle target recognition and achieve a good balance between time-consuming and accuracy.

show abstract

Section: Main Workmentioning

confidence: 99%

Robust LiDAR-Based Vehicle Detection for On-Road Autonomous Driving

Jin

Yang

et al. 2023

Remote Sensing

Self Cite

View full text Add to dashboard Cite

show abstract

“…Lee et al proposed a Geometric Model-Free Approach with a Particle Filter (GMFA-PF), which realized the good recognition and tracking of moving objects in the driving environment [29]. Jin et al designed a ground segmentation algorithm based on a Gaussian process, which effectively solved the problem of over-segmentation and under-segmentation in the traditional ground segmentation algorithm [30]. Lin et al proposed an adaptive clustering segmentation method to further segment the preliminary segmented point cloud according to the standard deviation of clustering points and realized the efficient segmentation of scene point cloud data [31].…”

Section: Introductionmentioning

confidence: 99%

Study on Obstacle Detection Method Based on Point Cloud Registration

Wang,

Wang

et al. 2024

WEVJ

View full text Add to dashboard Cite

An efficient obstacle detection system is one of the most important guarantees for improving the active safety performance of autonomous vehicles. This paper proposes an obstacle detection method based on high-precision positioning applied to blocked zones to solve the problems of the high complexity of detection results, low computational efficiency, and high load in traditional obstacle detection methods. Firstly, an NDT registration method which uses the likelihood function as the optimal value of the registration score function to calculate the registration parameters is designed to match the scanning point cloud and the target point cloud. Secondly, a target reduction method combined with threshold judgment and the binary tree search algorithm is designed to filter the point cloud of non-road obstacles to improve the processing speed of the computing platform. Meanwhile, KD-tree is used to speed up the clustering process. Finally, a vehicle remote control simulation platform with the combination of a cloud platform and mobile terminal is designed to verify the effectiveness of the strategy in practical application. The results prove that the proposed obstacle detection method can improve the efficiency and accuracy of detection.

show abstract

“…As autonomous driving technology evolves, lidar is becoming the key detection unit for the perception systems of autonomous vehicles, due to its ability to precisely detect range and angle information. The function of perception systems based on 3D lidar is to achieve obstacle detection from disordered point clouds: the obstacle detection results have been shown to directly influence the precision of motion planning and decision making [1][2][3][4][5][6]. Existing 3D point cloud processing methods are problematic, because they select improper clustering thresholds, which result in low detection accuracy; therefore, obstacle detection based on lidar is a promising prospect.…”

Section: Introductionmentioning

confidence: 99%

Obstacle Detection by Autonomous Vehicles: An Adaptive Neighborhood Search Radius Clustering Approach

et al. 2023

View full text Add to dashboard Cite

For autonomous vehicles, obstacle detection results using 3D lidar are in the form of point clouds, and are unevenly distributed in space. Clustering is a common means for point cloud processing; however, improper selection of clustering thresholds can lead to under-segmentation or over-segmentation of point clouds, resulting in false detection or missed detection of obstacles. In order to solve these problems, a new obstacle detection method was required. Firstly, we applied a distance-based filter and a ground segmentation algorithm, to pre-process the original 3D point cloud. Secondly, we proposed an adaptive neighborhood search radius clustering algorithm, based on the analysis of the relationship between the clustering radius and point cloud spatial distribution, adopting the point cloud pitch angle and the horizontal angle resolution of the lidar, to determine the clustering threshold. Finally, an autonomous vehicle platform and the offline autonomous driving KITTI dataset were used to conduct multi-scene comparative experiments between the proposed method and a Euclidean clustering method. The multi-scene real vehicle experimental results showed that our method improved clustering accuracy by 6.94%, and the KITTI dataset experimental results showed that the F1 score increased by 0.0629.

show abstract

A Robust Gaussian Process-Based LiDAR Ground Segmentation Algorithm for Autonomous Driving

Cited by 4 publications

References 28 publications

Robust LiDAR-Based Vehicle Detection for On-Road Autonomous Driving

Robust LiDAR-Based Vehicle Detection for On-Road Autonomous Driving

Study on Obstacle Detection Method Based on Point Cloud Registration

Obstacle Detection by Autonomous Vehicles: An Adaptive Neighborhood Search Radius Clustering Approach

Contact Info

Product

Resources

About