

浏览全部资源
扫码关注微信
1.南京信息工程大学 计算机学院、软件学院, 江苏 南京, 210044
2.无锡学院 物联网工程学院, 江苏 无锡, 214105
3.无锡学院 车路多模态感知与控制江苏省高校重点实验室, 江苏 无锡, 214105
Received:22 April 2026,
Revised:2026-05-16,
Published:06 June 2026
移动端阅览
CHEN Qidong, LI Jiawei, WU Hao, et al. Improved 3D object detection algorithm based on sparse enhancement and geometric attention[J]. Chinese Journal of Liquid Crystals and Displays, 2026, 41(6): 808-822.
CHEN Qidong, LI Jiawei, WU Hao, et al. Improved 3D object detection algorithm based on sparse enhancement and geometric attention[J]. Chinese Journal of Liquid Crystals and Displays, 2026, 41(6): 808-822. DOI: 10.37188/CJLCD.2026-0068. CSTR: 32172.14.CJLCD.2026-0068.
针对现有基于PointPillars的改进框架对行人等小目标检测精度不足的问题,提出一种融合稀疏增强与几何注意力的改进方法。首先,在Pillar特征编码阶段引入稀疏Pillar增强机制,填补稀疏区域的几何信息空白,为后续特征增强提供有效素材。其次,设计点感知空间注意力机制嵌入编码流程,基于补全后Pillar的点云密度与点到中心平均距离两类几何特征,通过轻量级MLP生成动态注意力权重,优先放大目标相关Pillar的特征信号,解决传统注意力依赖通道统计、不适配点云特性的缺陷。为验证方法的有效性,在公开KITTI三维目标检测数据集上开展实验,重点评估小目标的检测性能。实验结果表明,相较于现有的基于PointPillars的改进框架,三维模式下的行人目标的平均检测精度提升2.54%,鸟瞰图模式下的行人目标的平均检测精度提升5.91%,充分证明了所提创新模块对小目标特征优化的有效性,为自动驾驶长尾场景的感知任务提供了实用解决方案。
To address the issue of insufficient detection accuracy for small objects such as pedestrians in existing PointPillars-based frameworks, we propose an improved method that integrates sparse enhancement with geometric attention. First, we introduce a sparse Pillar enhancement mechanism during the Pillar feature encoding stage to fill in the gaps in geometric information in sparse regions, thereby providing effective material for subsequent feature enhancement. Second, we design a point-aware spatial attention mechanism embedded in the encoding process. Based on two types of geometric features of the completed Pillars, namely point cloud density and average point-to-center distance, we generate dynamic attention weights via a lightweight MLP to prioritize the amplification of feature signals from Pillars related to targets, thereby addressing the shortcomings of traditional attention mechanisms that rely on channel statistics and are ill-suited to point cloud characteristics. To validate the effectiveness of the method, experiments are conducted on the public KITTI 3D object detection dataset, with a focus on evaluating detection performance for small objects. The experimental results demonstrate that, compared to existing improved frameworks based on PointPillars, the average detection accuracy for pedestrian targets increases by 2.54% in 3D mode, and 5.91% in bird’s-eye-view mode. This results fully validates the effectiveness of the proposed innovative module in optimizing features for small targets and provides a practical solution for perception tasks in long-tail scenarios of autonomous driving.
ZHAO J Y , ZHAO W Y , DENG B , et al . Autonomous driving system: a comprehensive survey [J]. Expert Systems with Applications , 2024 , 242 : 122836 . doi: 10.1016/j.eswa.2023.122836 http://dx.doi.org/10.1016/j.eswa.2023.122836
秦静 , 王伟滨 , 邹启杰 , 等 . 基于激光雷达点云的3D目标检测方法综述 [J]. 计算机科学 , 2023 , 50 ( S1 ): 220400214 . doi: 10.11896/jsjkx.220400214 http://dx.doi.org/10.11896/jsjkx.220400214
QIN J , WANG W B , ZOU Q J , et al . Review of 3D target detection methods based on LiDAR point clouds [J]. Computer Science , 2023 , 50 ( S1 ): 220400214 . (in Chinese) . doi: 10.11896/jsjkx.220400214 http://dx.doi.org/10.11896/jsjkx.220400214
ZHANG L Y , ZHANG C , ZHANG L , et al . A dual-mode LiDAR system enabled by mechanically tunable hybrid cascaded metasurfaces [J]. Light: Science Applications , 2025 , 14 ( 1 ): 287 . doi: 10.1038/s41377-025-01999-4 http://dx.doi.org/10.1038/s41377-025-01999-4
YAO B C , WANG W T , XIE Z D , et al . Interdisciplinary advances in microcombs: bridging physics and information technology [J]. eLight , 2024 , 4 ( 1 ): 19 . doi: 10.1186/s43593-024-00071-9 http://dx.doi.org/10.1186/s43593-024-00071-9
SONG X N , ZHANG H Z , WANG L J , et al . Deep learning techniques for point cloud tasks: a review [J]. Applied Intelligence , 2025 , 55 ( 15 ): 991 . doi: 10.1007/s10489-025-06854-y http://dx.doi.org/10.1007/s10489-025-06854-y
ZHENG Y , WANG X , JIANG Z , et al . Adaptive multiscale microscope with fast zooming, extended working distance, and large field of view [J]. Light: Advanced Manufacturing , 2024 , 5 ( 1 ): 62 - 74 . doi: 10.37188/lam.2024.008 http://dx.doi.org/10.37188/lam.2024.008
孟毅儒 , 吕金光 , 郑凯丰 , 等 . 采用多角度偏振光学相干层析成像的滑油磨粒三维形貌检测和重建 [J]. 中国光学(中英文) , 2025 , 18 ( 6 ): 1449 - 1462 . doi: 10.37188/co.en-2025-0018 http://dx.doi.org/10.37188/co.en-2025-0018
MENG Y R , LV J G , ZHENG K F , et al . 3-D morphological feature measurement and reconstruction of wear particles using multi-view polarized optical coherence tomography [J]. Chinese Optics , 2025 , 18 ( 6 ): 1449 - 1462 . (in Chinese) . doi: 10.37188/co.en-2025-0018 http://dx.doi.org/10.37188/co.en-2025-0018
MATURANA D , SCHERER S . VoxNet: a 3D convolutional neural network for real-time object recognition [C]. 2015 IEEE/RSJ International Conference on Intelligent Robots and Systems . Hamburg, Germany : IEEE , 2015 : 922 - 928 . doi: 10.1109/iros.2015.7353481 http://dx.doi.org/10.1109/iros.2015.7353481
QI C R , HAO S , MO K C , et al . PointNet: deep learning on point sets for 3D classification and segmentation [C]. 2017 IEEE Conference on Computer Vision and Pattern Recognition . Honolulu, HI, USA : IEEE , 2017 : 77 - 85 . doi: 10.1109/cvpr.2017.16 http://dx.doi.org/10.1109/cvpr.2017.16
YAN Y , MAO Y X , LI B . SECOND: sparsely embedded convolutional detection [J]. Sensors , 2018 , 18 ( 10 ): 3337 . doi: 10.3390/s18103337 http://dx.doi.org/10.3390/s18103337
SHI S S , WANG X G , LI H S . PointRCNN: 3D object proposal generation and detection from point cloud [C]. 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition . Long Beach, CA, USA : IEEE , 2019 : 770 - 779 . doi: 10.1109/cvpr.2019.00086 http://dx.doi.org/10.1109/cvpr.2019.00086
ZHOU Y , TUZEL O . VoxelNet: end-to-end learning for point cloud based 3D object detection [C]// Proceedings of 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition . Salt Lake City, UT, USA : IEEE , 2018 : 4490 - 4499 . doi: 10.1109/cvpr.2018.00472 http://dx.doi.org/10.1109/cvpr.2018.00472
DENG J J , SHI S S , LI P W , et al . VoxeL R-CNN: towards high performance voxel-based 3D object detection [C]// Proceedings of the 35th AAAI Conference on Artificial Intelligence . Online : AAAI , 2021 : 1201 - 1209 . doi: 10.1609/aaai.v35i2.16207 http://dx.doi.org/10.1609/aaai.v35i2.16207
SHI S S , GUO C X , JIANG L , et al . PV-RCNN: point-voxel feature set abstraction for 3D object detection [C]. 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition . Seattle : IEEE , 2020 : 10526 - 10535 . doi: 10.1109/cvpr42600.2020.01054 http://dx.doi.org/10.1109/cvpr42600.2020.01054
LANG A H , VORA S , CAESAR H , et al . PointPillars: fast encoders for object detection from point clouds [C]. 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition . Long Beach, CA, USA : IEEE , 2019 : 12689 - 12697 . doi: 10.1109/cvpr.2019.01298 http://dx.doi.org/10.1109/cvpr.2019.01298
田枫 , 刘超 , 刘芳 , 等 . 基于改进PointPillars的激光雷达三维目标检测 [J]. 激光与光电子学进展 , 2024 , 61 ( 8 ): 0812007 . doi: 10.3788/LOP231493 http://dx.doi.org/10.3788/LOP231493
TIAN F , LIU C , LIU F , et al . Laser radar 3D target detection based on improved PointPillars [J]. Laser Optoelectronics Progress , 2024 , 61 ( 8 ): 0812007 . (in Chinese) . doi: 10.3788/LOP231493 http://dx.doi.org/10.3788/LOP231493
LIU Z , ZHAO X , HUANG T T , et al . TANet: robust 3D object detection from point clouds with triple attention [C]// Proceedings of the 34th AAAI Conference on Artificial Intelligence . New York, NY, USA : AAAI , 2020 : 11677 - 11684 . doi: 10.1609/aaai.v34i07.6837 http://dx.doi.org/10.1609/aaai.v34i07.6837
詹为钦 , 倪蓉蓉 , 杨彪 . 基于注意力机制的PointPillars+三维目标检测 [J]. 江苏大学学报(自然科学版) , 2020 , 41 ( 3 ): 268 - 273 .
ZHAN W Q , NI R R , YANG B . An attention-based PointPillars+3D object detection [J]. Journal of Jiangsu University (Natural Science Edition) , 2020 , 41 ( 3 ): 268 - 273 . (in Chinese)
LIU Y Y , YANG Z Y , TONG J L , et al . ET-PointPillars: improved PointPillars for 3D object detection based on optimized voxel downsampling [J]. Machine Vision and Applications , 2024 , 35 ( 3 ): 56 . doi: 10.1007/s00138-024-01538-y http://dx.doi.org/10.1007/s00138-024-01538-y
LI J Y , LUO C X , YANG X D . PillarNeXt: rethinking network designs for 3D object detection in LiDAR point clouds [C]// Proceedings of 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition . Vancouver, BC, Canada : IEEE , 2023 : 17567 - 17576 . doi: 10.1109/cvpr52729.2023.01685 http://dx.doi.org/10.1109/cvpr52729.2023.01685
高世严 , 柳杰 , 陈文艺 , 等 . 基于深度学习的单目标跟踪算法研究进展 [J]. 液晶与显示 , 2025 , 40 ( 8 ): 1202 - 1218 . doi: 10.37188/CJLCD.2025-0081 http://dx.doi.org/10.37188/CJLCD.2025-0081
GAO S Y , LIU J , CHEN W Y , et al . Review of single object tracking algorithm based on deep learning [J]. Chinese Journal of Liquid Crystals and Displays , 2025 , 40 ( 8 ): 1202 - 1218 . (in Chinese) . doi: 10.37188/CJLCD.2025-0081 http://dx.doi.org/10.37188/CJLCD.2025-0081
王伯霄 , 宋延嵩 , 董小娜 . 难点注意力感知红外小目标检测网络 [J]. 中国光学(中英文) , 2024 , 17 ( 3 ): 538 - 547 . doi: 10.37188/co.2023-0178 http://dx.doi.org/10.37188/co.2023-0178
WANG B X , SONG Y S , DONG X N . Indistinguishable points attention-aware network for infrared small object detection [J]. Chinese Optics , 2024 , 17 ( 3 ): 538 - 547 . (in Chinese) . doi: 10.37188/co.2023-0178 http://dx.doi.org/10.37188/co.2023-0178
SHU X N , ZHANG L . Research on PointPillars algorithm based on feature-enhanced backbone network [J]. Electronics , 2024 , 13 ( 7 ): 1233 . doi: 10.3390/electronics13071233 http://dx.doi.org/10.3390/electronics13071233
KONG W W , DU Y S , HE L L , et al . Improved 3D object detection based on pointpillars [J]. Electronics , 2024 , 13 ( 15 ): 2915 . doi: 10.3390/electronics13152915 http://dx.doi.org/10.3390/electronics13152915
张淦 , 彭育辉 , 孙宝哲 , 等 . 基于点云特征增强的PointPillars改进算法 [J]. 激光与光电子学进展 , 2025 , 62 ( 6 ): 0615012 . doi: 10.3788/LOP241730 http://dx.doi.org/10.3788/LOP241730
ZHANG G , PENG Y H , SUN B Z , et al . Improved PointPillars algorithm based on point cloud feature enhancement [J]. Laser Optoelectronics Progress , 2025 , 62 ( 6 ): 0615012 . (in Chinese) . doi: 10.3788/LOP241730 http://dx.doi.org/10.3788/LOP241730
GUAN X L , WAN H , HAN W K , et al . MDS-PointPillars: a lightweight obstacle identification method in farmland based on three-dimensional LiDAR for autonomous navigation [J]. Computers and Electronics in Agriculture , 2025 , 237 : 110688 . doi: 10.1016/j.compag.2025.110688 http://dx.doi.org/10.1016/j.compag.2025.110688
YAO X P , LIU P Y , ZHOU J M , et al . MAT-PointPillars: enhanced PointPillars algorithm based on multi-scale attention mechanisms and transformer [J]. PLoS One , 2025 , 20 ( 6 ): e0325373 . doi: 10.1371/journal.pone.0325373 http://dx.doi.org/10.1371/journal.pone.0325373
赵晶 , 李少博 , 郭杰龙 , 等 . 基于知识蒸馏和定位引导的Pointpillars点云检测网络 [J]. 液晶与显示 , 2024 , 39 ( 1 ): 79 - 88 . doi: 10.37188/cjlcd.2023-0058 http://dx.doi.org/10.37188/cjlcd.2023-0058
ZHAO J , LI S B , GUO J L , et al . Pointpillars point cloud detection network based on knowledge distillation and location guidance [J]. Chinese Journal of Liquid Crystals and Displays , 2024 , 39 ( 1 ): 79 - 88 . (in Chinese) . doi: 10.37188/cjlcd.2023-0058 http://dx.doi.org/10.37188/cjlcd.2023-0058
CHEN S T , ZHANG H L , ZHENG N N . Leveraging anchor-based LiDAR 3D object detection via point assisted sample selection [J]. IEEE Transactions on Intelligent Transportation Systems , 2025 , 26 ( 6 ): 7939 - 7952 . doi: 10.1109/tits.2025.3555229 http://dx.doi.org/10.1109/tits.2025.3555229
CHANDORKAR A , TERCAN H , MEISEN T . Rethinking backbone design for lightweight 3D object detection in LiDAR [C]// Proceedings of 2025 IEEE/CVF International Conference on Computer Vision . Honolulu, HI, USA : IEEE , 2025 : 1709 - 1717 . doi: 10.1109/iccvw69036.2025.00180 http://dx.doi.org/10.1109/iccvw69036.2025.00180
李佳男 , 王泽 , 许廷发 . 基于点云数据的三维目标检测技术研究进展 [J]. 光学学报 , 2023 , 43 ( 15 ): 1515001 . doi: 10.3788/AOS230745 http://dx.doi.org/10.3788/AOS230745
LI J N , WANG Z , XU T F . Three-dimensional object detection technology based on point cloud data [J]. Acta Optica Sinica , 2023 , 43 ( 15 ): 1515001 . (in Chinese) . doi: 10.3788/AOS230745 http://dx.doi.org/10.3788/AOS230745
SOHAIL S S , HIMEUR Y , KHEDDAR H , et al . Advancing 3D point cloud understanding through deep transfer learning: a comprehensive survey [J]. Information Fusion , 2025 , 113 : 102601 . doi: 10.1016/j.inffus.2024.102601 http://dx.doi.org/10.1016/j.inffus.2024.102601
缪永伟 , 刘家宗 , 陈佳慧 , 等 . 基于生成对抗网络的点云形状保结构补全 [J]. 中国科学:信息科学 , 2020 , 50 ( 5 ): 675 - 691 . doi: 10.1360/ssi-2019-0096 http://dx.doi.org/10.1360/ssi-2019-0096
MIAO Y W , LIU J Z , CHEN J H , et al . Structure-preserving shape completion of 3D point clouds with generative adversarial network [J]. Scientia Sinica Informationis , 2020 , 50 ( 5 ): 675 - 691 . (in Chinese) . doi: 10.1360/ssi-2019-0096 http://dx.doi.org/10.1360/ssi-2019-0096
潘李琳 , 邵剑飞 . 融合图注意力的多分辨率点云补全 [J]. 激光技术 , 2023 , 47 ( 5 ): 700 - 707 .
PAN L L , SHAO J F . Multi-resolution point cloud completion fusing graph attention [J]. Laser Technology , 2023 , 47 ( 5 ): 700 - 707 . (in Chinese)
ZHANG X S , HE Y , SONG C L , et al . Enhanced frustrum multi-scale VoteNet for 3D object detection in cluttered indoor scene [J]. Applied Intelligence , 2025 , 55 ( 7 ): 588 . doi: 10.1007/s10489-025-06492-4 http://dx.doi.org/10.1007/s10489-025-06492-4
SUN L B , LI Y F , QIN W H . PEPillar: a point-enhanced pillar network for efficient 3D object detection in autonomous driving [J]. The Visual Computer , 2025 , 41 ( 3 ): 1777 - 1788 . doi: 10.1007/s00371-024-03481-5 http://dx.doi.org/10.1007/s00371-024-03481-5
LEE M , PARK S , KIM H , et al . SPADE: sparse pillar-based 3d object detection accelerator for autonomous driving [C]. 2024 IEEE International Symposium on High-Performance Computer Architecture . Edinburgh, United Kingdom : IEEE , 2024 : 454 - 467 . doi: 10.1109/hpca57654.2024.00041 http://dx.doi.org/10.1109/hpca57654.2024.00041
PARK S , LEE M , CHOI J , et al . Selectively dilated convolution for accuracy-preserving sparse pillar-based embedded 3D object detection [C]// Proceedings of 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition . Seattle, WA, USA : IEEE , 2024 : 8104 - 8113 . doi: 10.1109/cvprw63382.2024.00810 http://dx.doi.org/10.1109/cvprw63382.2024.00810
WOO S , PARK J , LEE J Y , et al . CBAM: convolutional block attention module [C]// Proceedings of the 15th European Conference on Computer Vision . Munich, Germany : Springer , 2018 : 3 - 19 . doi: 10.1007/978-3-030-01234-2_1 http://dx.doi.org/10.1007/978-3-030-01234-2_1
BAJBAA K , USMAN M , ANWAR S , et al . Bird eye-view to street-view: a survey [J]. Applied Intelligence , 2025 , 55 ( 16 ): 1050 . doi: 10.1007/s10489-025-06940-1 http://dx.doi.org/10.1007/s10489-025-06940-1
GEIGER A , LENZ P , STILLER C , et al . Vision meets robotics: the KITTI dataset [J]. The International Journal of Robotics Research , 2013 , 32 ( 11 ): 1231 - 1237 . doi: 10.1177/0278364913491297 http://dx.doi.org/10.1177/0278364913491297
KINGMA D P , BA J . Adam: a method for stochastic optimization [C]// Proceedings of the 3rd International Conference on Learning Representations . San Diego, CA, USA : ICLR , 2015 .
GEIGER A , LENZ P , URTASUN R . Are we ready for autonomous driving? The KITTI vision benchmark suite [C]. 2012 IEEE Conference on Computer Vision and Pattern Recognition. Providence, RI, USA : IEEE , 2012 : 3354 - 3361 . doi: 10.1109/cvpr.2012.6248074 http://dx.doi.org/10.1109/cvpr.2012.6248074
0
Views
175
下载量
0
CSCD
Publicity Resources
Related Articles
Related Author
Related Institution
京公网安备11010802024621