基于自身运动约束的RGB-D动态SLAM算法

doi:10.3778/j.issn.1002-8331.2312-0362

摘要/Abstract

摘要： 针对视觉SLAM（simultaneous localization and mapping）算法在动态环境下容易出现重定位失败的问题，提出了一种基于自身运动约束的动态SLAM算法。采用YOLOv5s初步区分前景与背景特征点，仅利用背景特征点进行位姿初始化；利用IMU位姿信息不受动态环境影响的特性，计算每个特征点的自身运动约束值；根据背景特征点的特征约束结果设计了动态概率模型，自适应确定当前帧特征点约束的动态阈值，去除动态特征点并更新相机位姿。使用仿真数据集和真实环境数据集进行了实验验证。实验结果表明，在仿真数据集中，该方法能去除沿极线运动的特征点，相较于Dyna-SLAM，在均方根误差和标准差两项指标的提升率为87.81%和83.17%，相较于AirDos提升率分别为51.62%和41.91%。真实动态环境中重定位能力强于Dyna-SLAM，较AirDos无明显精度提升但运行速度提升29.48%。

关键词: 动态环境, 视觉SLAM, 多传感器融合, IMU预积分

Abstract: Aiming at the problem that visual SLAM algorithms are prone to repositioning failure in dynamic environments, a dynamic SLAM algorithm based on its own motion constraints is proposed. Firstly, YOLOv5s is used to initially distinguish foreground and background feature points, and only the background feature points are used for position initialization. Secondly, the IMU position information is not affected by the dynamic environment, and the own motion constraint value of each feature point is calculated. Finally, a dynamic probability model is designed based on the result of feature constraints on the background feature points, and the dynamic threshold of the feature point constraints of the current frame is adaptively determined to remove the dynamic feature points and update the camera position. Experimental validation is carried out using the simulation datasets and the real environment datasets. Experimental results show that in the simulation data set, this method can remove feature points that move along polar lines. Compared with Dyna-SLAM, the improvement rates in the root mean square error and standard deviation are 87.81% and 83.17%, respectively. Compared with AirDos, the improvement rates are 51.62% and 41.91% respectively. In the real dynamic environment, its repositioning ability is stronger than that of Dyna-SLAM and there is no obvious accuracy improvement but the running speed is increased by 29.48%.

Key words: dynamic scenes, visual SLAM, multi-sensor fusion, IMU pre-integration

张志铭, 张军, 刘元盛, 王子钰, 赵玉洁. 基于自身运动约束的RGB-D动态SLAM算法[J]. 计算机工程与应用, 2025, 61(8): 173-182.

ZHANG Zhiming, ZHANG Jun, LIU Yuansheng, WANG Ziyu, ZHAO Yujie. RGB-D Dynamic SLAM Algorithm Based on Own Motion Constraints[J]. Computer Engineering and Applications, 2025, 61(8): 173-182.

参考文献

[1] SAPUTRA M R U, MARKHAM A, TRIGONI N. Visual SLAM and structure from motion in dynamic environments: a survey[J]. ACM Computing Surveys, 2019, 51(2): 1-36.
[2] CAMPOS C, ELVIRA R, RODRIGUEZ J J G, et al. ORB-SLAM3: an accurate open-source library for visual, visual-inertial and multi-map SLAM[J]. IEEE Transactions on Robotics, 2021, 37(6): 1874-1890.
[3] KLEIN G, MURRAY D. Parallel tracking and mapping for small AR workspaces[C]//Proceedings of the 6th IEEE and ACM International Symposium on Mixed and Augmented Reality, 2007: 225-234.
[4] FORSTER C, PIZZOLI M, SCARAMUZZA D. SVO: fast semi-direct monocular visual odometry[C]//Proceedings of the IEEE International Conference on Robotics and Automation, 2014: 15-22.
[5] HARTLEY R, ZISSERMAN A. Multiple view geometry in computer vision[M]. Cambridge: Cambridge University Press, 2003.
[6] WU W, GUO L, GAO H, et al. YOLO-SLAM: a semantic SLAM system towards dynamic environment with geometric constraint[J]. Neural Computing and Applications, 2022, 34(8): 6011-6026.
[7] LI G H, CHEN S L. Visual SLAM in dynamic scenes based on object tracking and static points detection[J]. Journal of Intelligent &Robotic Systems, 2022, 104: 33.
[8] HU X, ZHANG Y Z, CAO Z Z, et al. CFP-SLAM: a real-time visual SLAM based on coarse-to-fine probability in dynamic environments[C]//Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems, 2022: 4399-4406.
[9] YU C, LIU Z, LIU X J, et al. DS-SLAM: a semantic visual SLAM towards dynamic environments[C]//Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems, 2018: 1168-1174.
[10] BESCOS B, FACIL J M, CIVERA J, et al. DynaSLAM: tracking, mapping and inpainting in dynamic scenes[J]. IEEE Robotics and Automation Letters, 2018, 3(4): 4076-4083.
[11] BESCOS B, CAMPOS C, TARDóS J D, et al. DynaSLAM II: tightly-coupled multi-object tracking and SLAM[J]. IEEE Robotics and Automation Letters, 2021, 6(3): 5191-5198.
[12] QIU Y, WANG C, WANG W, et al. AirDOS: dynamic SLAM benefits from articulated objects[C]//Proceedings of the International Conference on Robotics and Automation, 2022: 8047-8053.
[13] ZHUANG Y, JIA P, LIU Z, et al. Amos-SLAM: an anti-dynamics two-stage RGB-D SLAM approach[J]. IEEE Transactions on Instrumentation and Measurement, 2024, 73: 1-10.
[14] ZHU J, LI H, ZHANG T. Camera, LiDAR, and IMU based multi-sensor fusion SLAM: a survey[J]. Tsinghua Science and Technology, 2023, 29(2): 415-429.
[15] LIU J, LI X, LIU Y, et al. RGB-D inertial odometry for a resource-restricted robot in dynamic environments[J]. IEEE Robotics and Automation Letters, 2022, 7(4): 9573-9580.
[16] SONG S, LIM H, LEE A J, et al. DynaVINS: a visual-inertial slam for dynamic environments[J]. IEEE Robotics and Automation Letters, 2022, 7(4): 11523-11530.
[17] WONG Y S, LI C, NIESSNER M, et al. RigidFusion: RGB‐D scene reconstruction with rigidly‐moving objects[J]. Computer Graphics Forum, 2021, 40(2): 511-522.
[18] 郭瑞奇, 修睿, 孙勇, 等. 面向动态环境的紧耦合视觉惯性SLAM改进算法[J]. 计算机工程与应用, 2025, 61(4): 339-348.
GUO R Q, XIU R, SUN Y, et al. A improved tightly-coupled visual-inertial SLAM algorithm for dynamic environment[J]. Computer Engineering and Applications, 2025, 61(4): 339-348.
[19] 高贵, 伍宣衡, 王忠美, 等. V-SLAM深度学习闭环检测研究进展与展望[J]. 计算机工程与应用, 2022, 58(11): 47-59.
GAO G, WU X H, WANG Z M, et al. Research progress and prospect of V-SLAM deep learning loop closure detection[J]. Computer Engineering and Applications, 2022, 58(11): 47-59.
[20] YAN D, LI T, SHI C. Enhanced online calibration and initialization of visual-inertial SLAM system leveraging the structure information[J]. IEEE Transactions on Instrumentation and Measurement, 2023, 72: 1-15.
[21] QIN T, LI P, SHEN S. VINS-Mono: a robust and versatile monocular visual-inertial state estimator[J]. IEEE Transactions on Robotics, 2018, 34(4): 1004-1020.