Relation-aware place recognition network for large-scale point clouds

Abstract

Most existing point cloud-based place recognition methods emphasize feature extraction from individual points or local regions, while largely neglecting the relational information embedded within local neighborhoods. As a result, they often fail to capture discriminative relational patterns, leading to reduced recognition accuracy in scenes containing geometrically similar structures. In this paper, we propose a novel relation-aware network (RA-Net) for place recognition. RA-Net jointly exploits local relational cues and global contextual information to learn discriminative scene representations for large-scale point cloud-based place recognition. First, a spatial relation feature extraction (SRFE) module is proposed to exploit relational information embedded within local neighborhoods. By learning relation-aware weights and adaptively aggregating neighborhood information, the proposed module captures discriminative relational patterns by jointly considering feature discrepancies and spatial offsets. Furthermore, a global feature extraction (GFE) module is introduced to aggregate global feature statistics and integrate them with pointwise representations, enabling local features to be enhanced with global contextual cues. Experimental results on four benchmark datasets demonstrate that RA-Net can generate more discriminative global descriptors and achieve promising performance. It exhibits strong generalization capabilities for unseen scenes.

Keywords

Relation-aware / Point cloud / Place recognition / Deep learning

References

[1] Arandjelovic R , Gronat P , Torii A , et al., 2016. NetVLAD:CNN architecture for weakly supervised place recognition. Proc IEEE Conf on Computer Vision and Pattern Recognition, p.5297- 5307.

[2] Cattaneo D , Vaghi M , Valada A , 2022. LCDNet:deep loop closure detection and point cloud registration for LiDAR SLAM. IEEE Trans Robot, 38 (4): 2074- 2093.

[3] Chen LN , Wang H , Kong H , et al., 2023. Point-wise transformer with sparse convolution network for place recognition. IEEE Robot Autom Lett, 8 (6): 3414- 3421.

[4] Du J , Wang R , Cremers D , 2020. DH3D:deep hierarchical 3D descriptors for robust large-scale 6DoF relocalization. 16^th European Conf on Computer Vision, p.744- 762.

[5] Fan ZX , Song ZB , Liu HY , et al., 2022. SVT-Net:super light-weight sparse voxel transformer for large scale place recognition. Proc 36^th AAAI Conf on Artificial Intelligence, p.551- 560.

[6] Griffiths E , Haghighat M , Denman S , et al., 2025. HOTFormerLoc:hierarchical octree transformer for versatile LiDAR place recognition across ground and aerial views. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.6648- 6658.

[7] Hao W , Zhang WJ , Jin HY , 2024. SAGE-Net:employing spatial attention and geometric encoding for point cloud based place recognition. IEEE Robot Autom Lett, 9 (6): 4958- 4965.

[8] He L , Wang XL , Zhang H , 2016. M2DP:a novel 3D point cloud descriptor and its application in loop closure detection. IEEE/RSJ Int Conf on Intelligent Robots and Systems, p.231- 237.

[9] Hou ZX , Yan Y , Xu CZ , et al., 2022. HiTPR:hierarchical transformer for place recognition in point cloud. Int Conf on Robotics and Automation, p.2612- 2618.

[10] Hui L , Cheng MM , Xie J , et al., 2022. Efficient 3D point cloud feature learning for large-scale place recognition. IEEE Trans Image Process, 31: 1258- 1270.

[11] Kim G , Kim A , 2018. Scan context:egocentric spatial descriptor for place recognition within 3D point cloud map. IEEE/RSJ Int Conf on Intelligent Robots and Systems, p.4802- 4809.

[12] Komorowski J , 2021. MinkLoc3D:point cloud based large-scale place recognition. Proc IEEE Winter Conf on Applications of Computer Vision, p.1789- 1798.

[13] Komorowski J , 2022. Improving point cloud based place recognition with ranking-based loss and large batch training. 26^th Int Conf on Pattern Recognition, p.3699- 3705.

[14] Komorowski J , Wysoczańska M , Trzcinski T , 2021. MinkLoc++:LiDAR and monocular image fusion for place recognition. Int Joint Conf on Neural Networks, p.1- 8.

[15] Komorowski J , Wysoczańska M , Trzcinski T , 2022. EgoNN:egocentric neural network for point cloud based 6DoF relocalization at the city scale. IEEE Robot Autom Lett, 7 (2): 722- 729.

[16] Kong D , Li X , Ni PZ , et al., 2024. TopSPR-Net:topology aware segment-level point cloud learning descriptors for threedimensional place recognition in large-scale environments. IEEE Trans Ind Electron, 71 (10): 13406- 13416.

[17] Kong X , Yang XM , Zhai GY , et al., 2020. Semantic graph based place recognition for 3D point clouds. IEEE/RSJ Int Conf on Intelligent Robots and Systems, p.8216- 8223.

[18] Li QP , Zhuang Y , Huai JZ , et al., 2024. An efficient point cloud place recognition approach based on transformer in dynamic environment. ISPRS J Photogramm Remote Sens, 207: 14- 26.

[19] Liu CX , Chen GY , Song R , 2024. LPS-Net:lightweight parametershared network for point cloud-based place recognition. IEEE Int Conf on Robotics and Automation, p.448- 454.

[20] Liu Z , Zhou SB , Suo CZ , et al., 2019. LPD-Net:3D point cloud learning for large-scale place recognition and environment analysis. Proc IEEE/CVF Int Conf on Computer Vision, p.2831- 2840.

[21] Lu T , Wang LM , Wu GS , 2021. CGA-Net:category guided aggregation for point cloud semantic segmentation. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.11688- 11697.

[22] Ma JY , Zhang J , Xu JT , et al., 2022. OverlapTransformer:an efficient and yaw-angle-invariant transformer network for LiDARbased place recognition. IEEE Robot Autom Lett, 7 (3): 6958- 6965.

[23] Ma JY , Xiong GM , Xu JY , et al., 2024. CVTNet:a cross-view transformer network for LiDAR-based place recognition in autonomous driving environments. IEEE Trans Ind Inform, 20 (3): 4039- 4048.

[24] Maddern W , Pascoe G , Linegar C , et al., 2017. 1 year, 1000 km:the Oxford RobotCar dataset. Int J Robot Res, 36 (1): 3- 15.

[25] Miao Y , Engelmann F , Vysotska O , et al., 2025. SceneGraphLoc:cross-modal coarse visual localization on 3D scene graphs. 18^th European Conf on Computer Vision, p.127- 150.

[26] Qi CR , Su H , Mo K , et al., 2017. PointNet:deep learning on point sets for 3D classification and segmentation. Proc IEEE Conf on Computer Vision and Pattern Recognition, p.77- 85.

[27] Rizzini DL , 2017. Place recognition of 3D landmarks based on geometric relations. IEEE/RSJ Int Conf on Intelligent Robots and Systems, p.648- 654.

[28] Röhling T , Mack J , Schulz D , 2015. A fast histogram-based similarity measure for detecting loop closures in 3-D LIDAR data. IEEE/RSJ Int Conf on Intelligent Robots and Systems, p.736- 741.

[29] Rosinol A , Gupta A , Abate M , et al., 2020. 3D dynamic scene graphs:actionable spatial perception with places, objects, and humans. https://arxiv.org/abs/2002.06289

[30] Shu DW , Kwon J , 2024. Hierarchical bidirected graph convolutions for large-scale 3-D point cloud place recognition. IEEE Trans Neur Netw Learn Syst, 35 (7): 9651- 9662.

[31] Siva S , Nahman Z , Zhang H , 2020. Voxel-based representation learning for place recognition based on 3D point clouds. IEEE/RSJ Int Conf on Intelligent Robots and Systems, p.8351- 8357.

[32] Sun Q , Liu HY , He J , et al., 2020. DAGC:employing dual attention and graph convolution for point cloud based place recognition. Proc Int Conf on Multimedia Retrieval, p.224- 232.

[33] Tang WJ , Yang CG , An ZL , et al., 2026. EfficientVPR:toward efficient visual place recognition via scene-aware prompt tuning and adaptive feature enhancement. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.33738- 33748.

[34] Uy MA , Lee GH , 2018. PointNetVLAD:deep point cloud based retrieval for large-scale place recognition. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.4470- 4479.

[35] Vidanapathirana K , Ramezani M , Moghadam P , et al., 2022. LoGG3DNet:locally guided global descriptor learning for 3D place recognition. Int Conf on Robotics and Automation, p.2215- 2221.

[36] Wen CC , Huang H , Liu YS , 2023. Pyramid learnable tokens for 3D LiDAR place recognition. IEEE Int Conf on Robotics and Automation, p.4143- 4149.

[37] Xia Y , Xu YS , Li S , et al., 2021. SOE-Net:a self-attention and orientation encoding network for point cloud based place recognition. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.11343- 11352.

[38] Xia Y , Gladkova M , Wang R , et al., 2022. PVT3D:point voxel transformers for place recognition from sparse LiDAR scans. https://arxiv.org/pdf/2211.12542v1

[39] Xia Y , Gladkova M , Wang R , et al., 2023. CASSPR:cross attention single scan place recognition. Proc IEEE/CVF Int Conf on Computer Vision, p.8427- 8438.

[40] Zhang WX , Xiao CX , 2019. PCAN:3D attention map learning using contextual information for point cloud based retrieval. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.12428- 12437.

[41] Zhang YJ , Shi PC , Li JY , 2024. LiDAR-based place recognition for autonomous driving:a survey. ACM Comput Surv, 57 (4): 106.

[42] Zuo WL , Liu LG , Li YZ , et al., 2025. PRGS:patch-to-region graph search for visual place recognition. Pattern Recogn, 166: 111673.

Rights & permissions

The Authors. Published by Zhejiang University Press Co., Ltd.