SPS-RCNN: Semantic-Guided Proposal Sampling for 3D Object Detection from LiDAR Point Clouds

Three-dimensional object detection using LiDAR has attracted significant attention due to its resilience to lighting conditions and ability to capture detailed geometric information. However, existing methods still face challenges, such as a high proportion of background points in the sampled point...

Full description

Saved in:
Bibliographic Details
Main Authors: Hengxin Xu, Lei Yang, Shengya Zhao, Shan Tao, Xinran Tian, Kun Liu
Format: Article
Language:English
Published: MDPI AG 2025-02-01
Series:Sensors
Subjects:
Online Access:https://www.mdpi.com/1424-8220/25/4/1064
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850080955740454912
author Hengxin Xu
Lei Yang
Shengya Zhao
Shan Tao
Xinran Tian
Kun Liu
author_facet Hengxin Xu
Lei Yang
Shengya Zhao
Shan Tao
Xinran Tian
Kun Liu
author_sort Hengxin Xu
collection DOAJ
description Three-dimensional object detection using LiDAR has attracted significant attention due to its resilience to lighting conditions and ability to capture detailed geometric information. However, existing methods still face challenges, such as a high proportion of background points in the sampled point set and limited accuracy in detecting distant objects. To address these issues, we propose semantic-guided proposal sampling-RCNN (SPS-RCNN), a multi-stage detection framework based on point–voxel fusion. The framework comprises three components: a voxel-based region proposal network (RPN), a keypoint sampling stream (KSS), and a progressive refinement network (PRN). In the KSS, we propose a novel semantic-guided proposal sampling (SPS) method, which increases the proportion of foreground points and enhances sensitivity to outliers through multilevel sampling that integrates proposal-based local sampling and semantic-guided global sampling. In the PRN, a cascade attention module (CAM) is employed to aggregate features from multiple subnets, progressively refining region proposals to improve detection accuracy for medium- and long-range objects. Comprehensive experiments on the widely used KITTI dataset demonstrate that SPS-RCNN improves detection accuracy and exhibits enhanced robustness across categories compared to the baseline.
format Article
id doaj-art-6284c41e7082462a9262113fdcbd5e41
institution DOAJ
issn 1424-8220
language English
publishDate 2025-02-01
publisher MDPI AG
record_format Article
series Sensors
spelling doaj-art-6284c41e7082462a9262113fdcbd5e412025-08-20T02:44:50ZengMDPI AGSensors1424-82202025-02-01254106410.3390/s25041064SPS-RCNN: Semantic-Guided Proposal Sampling for 3D Object Detection from LiDAR Point CloudsHengxin Xu0Lei Yang1Shengya Zhao2Shan Tao3Xinran Tian4Kun Liu5College of Transportation, Shandong University of Science and Technology, Qingdao 266590, ChinaCollege of Ocean Science and Engineering, Shandong University of Science and Technology, Qingdao 266590, ChinaNational Deep Sea Center, Qingdao 266237, ChinaCollege of Ocean Science and Engineering, Shandong University of Science and Technology, Qingdao 266590, ChinaCollege of Ocean Science and Engineering, Shandong University of Science and Technology, Qingdao 266590, ChinaNational Deep Sea Center, Qingdao 266237, ChinaThree-dimensional object detection using LiDAR has attracted significant attention due to its resilience to lighting conditions and ability to capture detailed geometric information. However, existing methods still face challenges, such as a high proportion of background points in the sampled point set and limited accuracy in detecting distant objects. To address these issues, we propose semantic-guided proposal sampling-RCNN (SPS-RCNN), a multi-stage detection framework based on point–voxel fusion. The framework comprises three components: a voxel-based region proposal network (RPN), a keypoint sampling stream (KSS), and a progressive refinement network (PRN). In the KSS, we propose a novel semantic-guided proposal sampling (SPS) method, which increases the proportion of foreground points and enhances sensitivity to outliers through multilevel sampling that integrates proposal-based local sampling and semantic-guided global sampling. In the PRN, a cascade attention module (CAM) is employed to aggregate features from multiple subnets, progressively refining region proposals to improve detection accuracy for medium- and long-range objects. Comprehensive experiments on the widely used KITTI dataset demonstrate that SPS-RCNN improves detection accuracy and exhibits enhanced robustness across categories compared to the baseline.https://www.mdpi.com/1424-8220/25/4/10643D object detectionlight detection and ranging (lidar)point–voxel fusionsemantic-guided proposal samplingcascade network
spellingShingle Hengxin Xu
Lei Yang
Shengya Zhao
Shan Tao
Xinran Tian
Kun Liu
SPS-RCNN: Semantic-Guided Proposal Sampling for 3D Object Detection from LiDAR Point Clouds
Sensors
3D object detection
light detection and ranging (lidar)
point–voxel fusion
semantic-guided proposal sampling
cascade network
title SPS-RCNN: Semantic-Guided Proposal Sampling for 3D Object Detection from LiDAR Point Clouds
title_full SPS-RCNN: Semantic-Guided Proposal Sampling for 3D Object Detection from LiDAR Point Clouds
title_fullStr SPS-RCNN: Semantic-Guided Proposal Sampling for 3D Object Detection from LiDAR Point Clouds
title_full_unstemmed SPS-RCNN: Semantic-Guided Proposal Sampling for 3D Object Detection from LiDAR Point Clouds
title_short SPS-RCNN: Semantic-Guided Proposal Sampling for 3D Object Detection from LiDAR Point Clouds
title_sort sps rcnn semantic guided proposal sampling for 3d object detection from lidar point clouds
topic 3D object detection
light detection and ranging (lidar)
point–voxel fusion
semantic-guided proposal sampling
cascade network
url https://www.mdpi.com/1424-8220/25/4/1064
work_keys_str_mv AT hengxinxu spsrcnnsemanticguidedproposalsamplingfor3dobjectdetectionfromlidarpointclouds
AT leiyang spsrcnnsemanticguidedproposalsamplingfor3dobjectdetectionfromlidarpointclouds
AT shengyazhao spsrcnnsemanticguidedproposalsamplingfor3dobjectdetectionfromlidarpointclouds
AT shantao spsrcnnsemanticguidedproposalsamplingfor3dobjectdetectionfromlidarpointclouds
AT xinrantian spsrcnnsemanticguidedproposalsamplingfor3dobjectdetectionfromlidarpointclouds
AT kunliu spsrcnnsemanticguidedproposalsamplingfor3dobjectdetectionfromlidarpointclouds