基于航空图像与地面点云的跨模态地点识别方法

张志超; 徐友春; 陈晋生; 穆巍炜; 陆峰

doi:10.12382/bgxb.2024.1115

您当前的位置：

首页 >

文章列表页 >

基于航空图像与地面点云的跨模态地点识别方法

更新时间：2026-04-24

- 基于航空图像与地面点云的跨模态地点识别方法
- A Cross-modal Place Recognition Method Based on Aerial Images and Ground Point Clouds
- 兵工学报 2026年47卷第2期页码：241115
- 作者机构：
  
  陆军军事交通学院军事交通运输研究所，天津 300161
  空装驻天津地区第一军事代表室，天津 300385
- 作者简介：
  
  通信作者邮箱：xu56419@126.com
  通信作者邮箱：devil_cjs@163.com
- 基金信息：
- DOI：10.12382/bgxb.2024.1115
  中图分类号：
- 收稿：2024-12-11，
  
  网络首发：2025-12-25，
  
  纸质出版：2026-02-28
- 稿件说明：
移动端阅览
张志超, 徐友春, 陈晋生, 等. 基于航空图像与地面点云的跨模态地点识别方法[J]. 兵工学报, 2026,47(2):241115.

ZHANG Zhichao, XU Youchun, CHEN Jinsheng, et al. A Cross-modal Place Recognition Method Based on Aerial Images and Ground Point Clouds[J]. Acta Armamentarii, 2026, 47(2): 241115.
张志超, 徐友春, 陈晋生, 等. 基于航空图像与地面点云的跨模态地点识别方法[J]. 兵工学报, 2026,47(2):241115. DOI： 10.12382/bgxb.2024.1115.

ZHANG Zhichao, XU Youchun, CHEN Jinsheng, et al. A Cross-modal Place Recognition Method Based on Aerial Images and Ground Point Clouds[J]. Acta Armamentarii, 2026, 47(2): 241115. DOI： 10.12382/bgxb.2024.1115.

摘要

针对卫星拒止环境下无人车在未知区域自主定位难题，提出一种从航空图像到地面点云的跨模态地点识别方法，并设计相应的网络架构（Aerial-to-Ground Position Recognition Network，AG-PRNet）。该方法通过数据预处理将点云投影到鸟瞰视图（Bird's Eye View，BEV）空间，减小其与航空图像的模态差异；设计旋转平移不变特征编码模块（Rotation And Translation Invariant CNN，RATI-CNN），提取跨模态数据的旋转平移不变特征；利用交叉注意力模块融合学习跨模态数据的共享特征，提升特征匹配的鲁棒性。在自建跨网域地点识别（ Cross-Domain Place Recognition，CDPR）数据集上的实验表明，所提方法的Top-1和Top-5召回率分别达60.08%和76%，显著优于基线方法，验证了其在跨模态地点识别中的有效性。

Abstract

To address the challenges of autonomous localization of unmanned vehicles in unknown areas in satellite-denied environments

this paper proposes a cross-modal place recognition method based on aerial images and ground point clouds and designs an aerial-to-ground position recognition network(AG-PRNet)architecture. The proposed method reduces the modal differences among point clouds and aerial images by projecting the point clouds into the bird's eye view(BEV)space through data preprocessing. A rotation and translation invariant convolutional neural network(RATI-CNN)module is designed to extract the invariant geometric features from cross-modal data. Additionally

a cross-attention module is employed to fuse the shared features of cross-modal data

significantly enhancing the robustness of feature matching. Experimental results on a self-constructed cross-domain place recognition(CDPR)dataset demonstrate that the proposed method achieves a Top-1 recall rate of 60.08% and a Top-5 recall rate of 76%

outperforming baseline methods and validating its effectiveness in cross-modal place recognition.

关键词

Keywords

references

李康宇,王西峰,朱守泰.基于局部与全局描述符紧耦合联合决策的机器人分层式视觉位置识别方法[J].信息与控制, 2024, 53(3): 400-415.

LI K Y, WANG X F, ZHU S T. Ahierarchical visual place recognition method for robots based on tightly-coupled joint decision of local and global descriptors[J]. Information and Control, 2024, 53(3): 400-415. (in Chinese)

LUO K, YU H S, CHEN X Y L, et al. 3D point cloud-based place recognition: A survey[J]. Artificial Intelligence Review, 2024, 57(4): 83-127.

DU Z H, JI S P, KHOSHELHAM K. 3D LiDAR-based place recognition techniques: A review of the past 10 years[J]. IEEE Transactions on Instrumentation and Measurement, 2024, 73: 1-24.

YIN H, XU X C. Asurvey on global LiDAR localization:challenges, advances and open problems[J]. International Journal of Computer Vision, 2024, 132(8): 3139-3171.

CHEN L, FENG C, MA Y. A review of rigid point cloud registration based on deep learning[J]. Frontiers in Neurorobotics, 2024, 17: 1-22.

张婉怡,王佳,宋明星.金字塔局部聚合描述符的视觉位置识别研究[J].测试技术学报,2025,39(1): 54-62.

ZHANG W Y, WANG J, SONG M X. Research on visual place recognition based on pyramid local aggregation descriptors[J]. Journal of Test Technology, 2025, 39(1): 54-62. (in Chinese)

SHIY J, LI H D, et al. Beyond cross-view image retrieval:Highly accurate vehicle localization using satellite image[C]∥Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New Orleans, LA, US: IEEE, 2022: 16989-16999.

张正华,陈国良.一种轻量且旋转不变的激光雷达位置识别网络[J].测绘学报,2025,(1): 90-103.

ZHANG Z H, CHEN G L. A Lightweight and rotation-invariant LiDAR place recognition network[J]. Acta Geodaetica et Cartographica Sinica, 2025, (1) : 90-103. (in Chinese)

TANG T Y, DE MARTINI D, NEWMAN P. Get to the Point:Learning LiDAR Place Recognition and Metric Localization Using Overhead Imagery[C]∥Proceedings of Robotics: Science and Systems XVII, Virtual Conference, 2021: 1-10.

TANG T Y, DE MARTINI D. Point-based metric and topological localisation between LiDAR and overhead imagery[J]. Autonomous Robots, 2023, 47(5): 595-615.

HU D I, YUAN X, XI H Y. Road structure inspired UGV-satellite cross-view geo-localization[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2024, 17: 16767-16786.

CHO Y, KIM G, LEE S, et al. OpenStreetMap-based lidar global localization in urban environment without a prior lidar map[J]. IEEE Robotics and Automation Letters, 2022, 7(2): 4999-5006.

GUAN T R, XIAN R, WANG X, et al. AGL-Net: Aerial-ground cross-, modal global localization with varying scales[C]∥Proceedings of the 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems. Kyoto, Japan: IEEE, 2024:8161-8169.

MING Y H, XU M Y, WANG X R, et al. VIPeR: Visual incremental place recognition with adaptive mining and continual learning[J]. IEEE Robotics and Automation Letters, 2025, 10 (3): 3038-3045.

LUO L, ZHENG S H, LI Y X, et al. BEVPlace: Learning LiDAR-based place recognition using bird's eye view images[C]∥Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV 2023). Paris, France: IEEE, 2023:8700-8709.

FAN Y Z, DU X, LUO L, et al. Fresco: Frequency-domain scan context for LiDAR-based place recognition with translation and rotation invariance[C]∥ Proceedings of the 2022 17th International Conference on Control, Automation, Robotics and Vision. Singapore: IEEE, 2022,pp: 576-583.

COHEN T, WELLING M. Group equivariant convolutional networks[C]∥Proceedings of the 33rd International Conference on Machine Learning. New York, NY, US: PMLR, 2016:2990-2999.

吴鹏翔,李凡长.一种基于群等变卷积的度量元学习算法[J].计算机工程, 2022, 48(2): 72-78.

WUPX,LIFC. Ametric meta-learning algorithm based on group equivariant convolution[J]. Computer Engineering,2022,48(2): 72-78. (in Chinese)

MO H L, ZHAO G Y. RIC-CNN: Rotation-invariant coordinate convolutional neural network[J]. Pattern Recognition, 2022, 146: 109994.

肖锋,卢浩,张文娟,等.基于旋转等变卷积的航拍红外图像目标识别算法[J].兵工学报,2024,45(8):2817-2827.

XIAO F, LU H, ZHANG W J, et al. Aerial infrared image target recognition algorithm based on rotation equivariant convolution[J]. Acta Armamentarii, 2024, 45(8): 2817-2827. (in Chinese)

CUI S, MA A L, ZHANG L P, et al. MAP-Net: SAR and optical image matching via image-based convolutional network with attention mechanism and spatial pyramid aggregated pooling[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 1000513.

YAO F L, LIU N Y, LI P G, et al. Cross-modal remote sensing image retrieval via intra-and inter-modal feature matching[C]∥Proceedings of the IEEE International Geoscience and Remote Sensing Symposium. Kuala Lumpur, Malaysia: IEEE, 2022,pp:1792-1795.

ZHANG Y X, LAN C Z, ZHANG H M, et al. Multimodal pemote sensing image matching via learning features and attention mechanism[J]. IEEE Transactions on Geoscience and Remote Sensing, 2024, 62: 5603620.

施成浩,陈谢沅澧,郭瑞斌,等.面向点云配准和地点识别的多头旋转注意力网络[J].控制理论与应用, 2023, 40(12): 2187-2197.

SHI CH,CHENXYL,GUORB, et al. A multi-head rotation attention network for point cloud registration and place recognition[J]. Control Theory & Applications, 2023, 40(12): 2187-2197. (in Chinese)

ARANDJELOVIˊC R, GRONAT P, TORII A. NetVLAD: CNN architecture for weakly supervised place recognition[C]∥Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas, NV, US: IEEE, 2016: 5297-5307.

KOMOROWSKI J. MinkLoc3D: Point-cloud-based large-scale place recognition[C]∥Proceedings of the 2021 IEEE Winter Conference on Applications of Computer Vision. Waikoloa, HI, US: IEEE, 2021: 1789-1798.

MA J Y, XIONG G M, XU J X, et al. CVTNet: A cross-view transformer network for LiDAR-based place recognition in autonomous driving environments[J]. IEEE Transactions on Industrial Informatics, 2024, 20(3): 4039-4048.

浏览量

下载量

CNKI被引量

文章被引用时，请邮件提醒。

提交

工具集

关联资源

暂无数据