基于大场景下识别动态物体的视觉SLAM研究
Visual SLAM Research Based on Recognizing Dynamic Objects in Large Scenes
DOI: 10.12677/mos.2024.134379, PDF,    国家自然科学基金支持
作者: 李兴州, 何 锋*：贵州大学机械工程学院，贵州 贵阳；余国宽：贵州师范大学机械与电气工程学院，贵州 贵阳
关键词: 视觉SLAM；动态场景；YOLOv7；动态目标点；Visual SLAM； Dynamic Scenes； YOLOv7； Dynamic Target Points
摘要: 目前轮式机器人基于YOLOv5的视觉SLAM算法并没有融合IMU解决大场景下鲁棒性较差的问题。利用目前较为先进的YOLOv7动态目标检测算法，将其与ORB_SLAM3算法的IMU数据融合，根据其检测的特征点判断是否为动态目标，保证其在动态场景下稳定工作。通过仿真分析，在TUM-VI数据集的动态场景下剔除动态目标特征点稳定，在Mono-IMU和Stereo-IMU两种模式下绝对误差精度平均提高30%以上，相对误差精度平均提升20%以上。因此，本文所采用的方法在大场景且存在动态物体的情况下提高了定位与建图精度。
Abstract: The current YOLOv5 vision-based SLAM algorithm for wheeled robots does not fuse IMU to solve the problem of poor robustness in large scenes. Using the more advanced YOLOv7 dynamic target detection algorithm, it fuses the data with the IMU data of the ORB_SLAM3 algorithm to determine whether it is a dynamic target or not according to its detected feature points, which ensures that it can work stably in dynamic scenes. Through simulation analysis, the rejection of dynamic target feature points is stabilized under the dynamic scene of the TUM-VI dataset, and the absolute error accuracy is improved by more than 30% on average and the relative error accuracy is improved by more than 20% on average in both Mono-IMU and Stereo-IMU modes. Therefore, the method adopted in this paper improves the localization and map building accuracy in large scenes with the presence of dynamic objects.
文章引用：李兴州, 何锋, 余国宽. 基于大场景下识别动态物体的视觉SLAM研究[J]. 建模与仿真, 2024, 13(4): 4180-4194. https://doi.org/10.12677/mos.2024.134379

参考文献

[1] 程俊廷, 郭博洋, 田宽. 改进的LK光流法在SLAM中的应用[J]. 黑龙江科技大学学报, 2019, 29(6): 736-740.
[2] 杨永刚, 武楚健, 杨正全. 基于融合改进RANSAC光流法的无人机视觉SLAM研究[J]. 半导体光电, 2023, 44(2): 277-283.
[3] Kundu, A., Krishna, K.M. and Sivaswamy, J. (2009) Moving Object Detection by Multi-View Geometric Techniques from a Single Camera Mounted Robot. 2009 IEEE/RSJ International Conference on Intelligent Robots and Systems, St. Louis, 10-15 October 2009, 4306-4312. [Google Scholar] [CrossRef] 
[4] Wang, R., Wan, W., Wang, Y. and Di, K. (2019) A New RGB-D SLAM Method with Moving Object Detection for Dynamic Indoor Scenes. Remote Sensing, 11, Article 1143. [Google Scholar] [CrossRef] 
[5] Fang, Y., Dai, B., et al. (2009) An Improved Moving Target Detecting and Tracking Based on Optical Flow Technique and Kalman Filter. 2009 4th International Conference on Computer Science & Education, Nanning, 25-28 July 2009, 1197-1202. [Google Scholar] [CrossRef] 
[6] Wang, Y. and Huang, S. (2014) Towards Dense Moving Object Segmentation Based Robust Dense RGB-D SLAM in Dynamic Scenarios. 2014 13th International Conference on Control Automation Robotics & Vision (ICARCV), Singapore, 10-12 December 2014, 1841-1846. [Google Scholar] [CrossRef] 
[7] Sun, Y., Liu, M. and Meng, M.Q.-H. (2017) Improving RGB-D SLAM in Dynamic Environments: A Motion Removal Approach. Robotics and Autonomous Systems, 89, 110-122. [Google Scholar] [CrossRef] 
[8] Xu, B., Li, W., Tzoumanikas, D., Bloesch, M., Davison, A. and Leutenegger, S. (2019) MID-Fusion: Octree-Based Object-Level Multi-Instance Dynamic SLAM. 2019 International Conference on Robotics and Automation (ICRA), Montreal, 20-24 May 2019, 5231-5237. [Google Scholar] [CrossRef] 
[9] Long, X., Zhang, W. and Zhao, B. (2020) PSPNet-SLAM: A Semantic SLAM Detect Dynamic Object by Pyramid Scene Parsing Network. IEEE Access, 8, 214685-214695. [Google Scholar] [CrossRef] 
[10] Ji, T., Wang, C. and Xie, L. (2021) Towards Real-Time Semantic RGB-D SLAM in Dynamic Environments. 2021 IEEE International Conference on Robotics and Automation (ICRA), Xi’an, 30 May-05 June 2021, 11175-11181. [Google Scholar] [CrossRef] 
[11] Zhang, H., Zhang, T., Lam, T.L. and Vijayakumar, S. (2021) PoseFusion2: Simultaneous Background Reconstruction and Human Shape Recovery in Real-time. 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Prague, 27 September-01 October 2021, 7631-7638. [Google Scholar] [CrossRef] 
[12] Zhang, T., Zhang, H., Li, X., Chen, J., Lam, T.L. and Vijayakumar, S. (2021) Acousticfusion: Fusing Sound Source Localization to Visual SLAM in Dynamic Environments. 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Prague, 27 September-1 October 2021, 6868-6875. [Google Scholar] [CrossRef] 
[13] Hu, X., Zhang, Y., Cao, Z., Ma, R., Wu, Y., Deng, Z., et al. (2022) CFP-SLAM: A Real-Time Visual SLAM Based on Coarse-to-Fine Probability in Dynamic Environments. 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Kyoto, 23-27 October 2022, 4399-4406. [Google Scholar] [CrossRef] 
[14] Xiao, L., Wang, J., Qiu, X., Rong, Z. and Zou, X. (2019) Dynamic-SLAM: Semantic Monocular Visual Localization and Mapping Based on Deep Learning in Dynamic Environment. Robotics and Autonomous Systems, 117, 1-16. [Google Scholar] [CrossRef] 
[15] Mur-Artal, R. and Tardos, J.D. (2017) ORB-SLAM2: An Open-Source SLAM System for Monocular, Stereo, and RGB-D Cameras. IEEE Transactions on Robotics, 33, 1255-1262. [Google Scholar] [CrossRef] 
[16] Zhang, R. and Zhang, X. (2023) Geometric Constraint-Based and Improved YOLOv5 Semantic SLAM for Dynamic Scenes. ISPRS International Journal of Geo-Information, 12, Article 211. [Google Scholar] [CrossRef] 
[17] Mseddi, W.S., Sedrine, M.A. and Attia, R. (2021) YOLOv5 Based Visual Localization for Autonomous Vehicles. 2021 29th European Signal Processing Conference (EUSIPCO), Dublin, 23-27 August 2021, 746-750. [Google Scholar] [CrossRef] 
[18] Wang, C., Bochkovskiy, A. and Liao, H.M. (2023) YOLOv7: Trainable Bag-of-Freebies Sets New State-of-the-Art for Real-Time Object Detectors. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Vancouver, 17-24 June 2023, 7464-7475. [Google Scholar] [CrossRef] 
[19] Campos, C., Elvira, R., Rodriguez, J.J.G., M. Montiel, J.M. and D. Tardos, J. (2021) ORB-SLAM3: An Accurate Open-Source Library for Visual, Visual-Inertial, and Multimap SLAM. IEEE Transactions on Robotics, 37, 1874-1890. [Google Scholar] [CrossRef] 
[20] Zhang, Z. and Scaramuzza, D. (2018) A Tutorial on Quantitative Trajectory Evaluation for Visual (-Inertial) Odometry. 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Madrid, 1-5 October 2018, 7244-7251. [Google Scholar] [CrossRef] 

为你推荐