School of Computer Science and Technology, Shandong University of Technology, Zibo 255000, China
Show less
文章历史+
Received
Published
Revised
2025-09-15
2026-06-18
2025-11-27
Online Date
2026-08-04
PDF (12497K)
摘要
为解决室内视觉定位中因家具、行人等遮挡物体导致的定位精度下降问题,提出了一种基于遮挡清除算法的室内视觉定位系统。系统首先利用YOLOv11目标检测算法识别并定位遮挡物体,对遮挡区域像素进行重置并生成掩码图像;随后采用MAT(mask-aware transformer)图像修复算法恢复无遮挡图像,以获取完整的环境信息用于后续定位。在定位阶段,基于本质矩阵的定位算法结合SuperPoint特征提取与LightGlue 特征匹配方法,有效提升了在光照变化和视点偏移下的鲁棒性。实验结果表明,提出的系统在 7Scenes数据集和自制LH数据集上表现优异。在7Scenes上,定位误差为(0.063 m, 1.759°),相比室内视觉定位领域最广泛应用于工业上SuperPoint + SuperGlue算法,位置精度提升约6%,旋转误差降低约4%;在LH数据集上,位置精度进一步提升约15%。在与最先进的算法对比时,尽管精度略逊一筹,但考虑到该方法使用的是去遮挡后的图像数据集,其在复杂场景中的表现依然出色,尤其在处理遮挡问题时能够保持较高的鲁棒性和精度。此外,该方法更具轻量性,能够更快速地部署和训练,适用于实时应用场景。
Abstract
In order to address the issue of decreased localization accuracy in indoor visual localization caused by occluding objects such as furniture and pedestrians, an indoor visual localization system based on an occlusion removal algorithm was proposed. Firstly, the YOLOv11 object detection algorithm was used to identify and locate occluded objects, reset the pixels in the occluded area, and generate a masked image. Subsequently, the mask aware transformer(MAT) image restoration algorithm was used to restore the unobstructed image, in order to obtain complete environmental information for subsequent localization. During the localization phase, the essential matrix-based localization algorithm combined with SuperPoint feature extraction and LightGlue feature matching methods effectively improves robustness under lighting changes and viewpoint shifts. Experimental results show that the proposed system performs excellently on both the 7Scenes dataset and the custom LH dataset. On the 7Scenes dataset, the localization error is (0.063 m, 1.759°), which improves position accuracy by approximately 6% and reduces rotation error by about 4% compared to the widely used SuperPoint + SuperGlue algorithm in the indoor visual localization field. On the LH dataset, position accuracy is further improved by around 15%. When compared to the state-of-the-art algorithms, although the accuracy is slightly lower, the proposed method, which uses occlusion-free image datasets, still performs excellently in complex scenes, especially in handling occlusions, maintaining high robustness and precision. Additionally, the proposed method is more lightweight, enabling faster deployment and training, making it suitable for real-time application scenarios.
ZHANG Li-ye, HE Fa-xu, CHENG Yi-shuo, CHENG Xin-ru, LI Peng-fei.
Indoor Visual Localization System Based on Occlusion Removal Algorithm[J].
Science Technology and Engineering, 2026, 26(17): 7352-7364 DOI:10.12404/j.issn.1671-1815.2506649
然后,提取查询图像特征,生成查询特征文件,并与数据库特征文件进行图像检索,找到最匹配的数据库图像。随后,通过特征匹配建立查询与数据库图像的2D-2D匹配信息。最后,基于本质矩阵的定位算法,利用RANSAC-五点法求解本质矩阵,并通过奇异值分解(singular value decomposition, SVD)优化,最终根据数据库图像的绝对位姿计算查询图像的定位结果。
GuoX, QinD, YangY. Indoor visual positioning based on image retrieval in dense connected convolutional network[C]// International Conference on Artificial Intelligence for Communications and Networks. Cham: Springer International Publishing, 2020: 70-79.
[2]
FischerG K J, SchaechtleT, GabbrielliA, et al. A systematic survey and comparative analysis of angular-based indoor localization and positioning technologies[J]. IEEE Communications Surveys & Tutorials, 2026, 28: 3830-3869.
[3]
ZhaoZ, HeW, LiF, et al.Enhancing indoor positioning with acoustic signal-aided visual-inertial SLAM[C]//2024 5th International Conference on Computer Engineering and Application (ICCEA). Hangzhou: IEEE, 2024: 1156-1159.
[4]
WuX, MiaoY, SunZ. ORB-YOLO:an indoor IMU-aided visual-inertial SLAM system for dynamic environment[C]//2023 International Conference on Artificial Intelligence of Things and Systems (AIoTSys). Xi’an: IEEE, 2023: 71-78.
[5]
QinP, HuQ, YuH. An Internet of Electronic-Visual Things indoor localization system using adaptive Kalman filter[J]. IEEE Sensors Journal, 2023, 23(14): 16058-16067.
[6]
WangJ, QiuJ, HuN.Robust 2D visual localization under indoor scenes[C]//2023 7th Asian Conference on Artificial Intelligence Technology (ACAIT). Jiaxing: IEEE, 2023: 1281-1288.
[7]
LaskarZ, MelekhovI, KaliaS, et al. Camera relocalization by computing pairwise relative poses using convolutional neural network[C]// Proceedings of the IEEE international conference on computer vision workshops. Venice: IEEE, 2017: 929-938.
[8]
ZhouQ, SattlerT, PollefeysM, et al. To learn or not to learn:visual localization from essential matrices[C]//2020 IEEE International Conference on Robotics and Automation (ICRA). Paris: IEEE, 2020: 3319-3326.
LüMengyan, GuoXinxin, ZhangKunpeng, et al. A visual indoor localization method based on fast image deblurring[J]. Information Technology and Informatization, 2024(4): 15-19.
[11]
LindenbergerP, SarlinP E, PollefeysM. Lightglue: local feature matching at light speed[C]// Proceedings of the IEEE/CVF International Conference on Computer Vision. Paris: IEEE, 2023: 17627-17638.
[12]
YangS, MaL. Diffusion model-based pedestrian de-occlusion method using a monocular camera sensor for indoor visual localization[J]. IEEE Sensors Journal, 2023, 23(22): 28421-28429.
XieTing, ZhangXiaojie, YeZhichao, et al. Visual localization system of integrated active and passive perception for indoor scenes[J]. Journal of Image and Graphics, 2023, 28(2): 522-534.
[15]
TairaH, OkutomiM, SattlerT, et al. InLoc: indoor visual localization with dense matching and view synthesis[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 43(4): 1293-1307.
[16]
DingY, LiuM, ZhangM, et al. Removing visual occlusion of construction scaffolds via a two-step method combining semantic segmentation and image inpainting[J]. Engineering Applications of Artificial Intelligence, 2025, 142: 109983.
[17]
HuY, YeS, ZhaoW, et al. O2-Recon: completing 3D reconstruction of occluded objects in the scene with a pre-trained 2D diffusion model[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2024, 38(3): 2285-2293.
PengHong, ZhouBin, SunYan, et al. Vector-valued PDE-constrained image inpainting model[J]. Science Technology and Engineering, 2025, 25(5): 2019-2026.
[20]
LiW, LinZ, ZhouK, et al. MAT: mask-aware transformer for large hole image inpainting[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New Orleans, LA: IEEE, 2022: 10758-10768.
[21]
DeshmukhR A, HasamnisM A, KulkarniM B, et al. Advancing indoor positioning systems: innovations, challenges, and applications in mobile robotics[J]. Robotica, 2025, 43(7): 1-41.
LinT Y, MaireM, BelongieS, et al. Microsoft coco: Common objects in context[C]// European Conference on Computer Vision. Cham: Springer International Publishing, 2014: 740-755.
[27]
ShottonJ, GlockerB, ZachC, et al. Scene coordinate regression forests for camera relocalization in RGB-D images[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Portland, OR: IEEE, 2013: 2930-2937.
[28]
ArandjelovicR, ZissermanA. All about VLAD[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Portland, OR: IEEE, 2013: 1578-1585.
[29]
KendallA, GrimesM, CipollaR. Convolutional networks for real-time 6-dof camera relocalization[J]. arXiv Preprint, arXiv: 1505. 07427, 2015.
[30]
KendallA, CipollaR. Geometric loss functions for camera pose regression with deep learning[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Honolulu, HI: IEEE, 2017: 5974-5983.
[31]
KendallA, CipollaR.Modelling uncertainty in deep learning for camera relocalization[C]//2016 IEEE international conference on Robotics and Automation (ICRA). Stockholm: IEEE, 2016: 4762-4769.
[32]
WalchF, HazirbasC, Leal-TaixeL, et al. Image-based localization using LSTMs for structured feature correlation[C]// Proceedings of the IEEE International Conference on Computer Vision. Venice: IEEE, 2017: 627-637.
[33]
BrahmbhattS, GuJ, KimK, et al. Geometry-aware learning of maps for camera localization[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Salt Lake City, UT: IEEE, 2018: 2616-2625.
[34]
XueF, WuX, CaiS, et al. Learning multi-view camera relocalization with graph neural networks[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Seattle, WA: IEEE, 2020: 11372-11381.
[35]
WangB, ChenC, LuC X, et al. AtLoc: attention guided camera localization[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2020, 34(6): 10393-10401.
[36]
TianM, NieQ, ShenH.3D scene geometry-aware constraint for camera localization with deep learning[C]//2020 IEEE international conference on robotics and automation (ICRA). Paris: IEEE, 2020: 4211-4217.
[37]
ChenS, LiX, WangZ, et al. DFNet: enhance absolute pose regression with direct feature matching[C]// European Conference on Computer Vision. Cham: Springer Nature Switzerland, 2022: 1-17.
[38]
TangS, LiY, WanJ, et al. TransCNNLoc: end-to-end pixel-level learning for 2D-to-3D pose estimation in dynamic indoor scenes[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2024, 207: 218-230.
[39]
BalntasV, LiS, PrisacariuV. RelocNet: Continuous metric learning relocalisation using neural nets[C]// Proceedings of the European Conference on Computer Vision (ECCV). Cham: Springer Nature Switzerland, 2018: 751-767.
[40]
SahaS, VarmaG, JawaharC V. Improved visual relocalization by discovering anchor points[J]. arXiv Preprint, arXiv: 1811. 04370, 2018.
[41]
DaiK, JiangZ, QiuF, et al. HVLF: a holistic visual localization framework across diverse scenes[J]. IEEE Transactions on Neural Networks and Learning Systems, 2025, 36(10): 18859-18873.