面向野外环境的视觉惯导SLAM算法  

Visual-inertial SLAM algorithm for off-road environments

作　　者：李静博 伊克萨尼·普尔凯提 朱斌 朱纪洪 艾斯卡尔·艾木都拉[1] LI Jing-bo;Eksan·Firkat;ZHU Bin;ZHU Ji-hong;Askar·Hamdulla(School of Computer Science and Technology,Xinjiang University,Xinjiang 830017,China;Department of Automation,Tsinghua University,Beijing 100084,China;Department of Precision Instrument,Tsinghua University,Beijing 100084,China)

机构地区：[1]新疆大学计算机科学与技术学院,新疆乌鲁木齐830017 [2]清华大学自动化系,北京100084 [3]清华大学精密仪器系,北京100084

出　　处：《计算机工程与设计》2025年第4期1005-1012,共8页Computer Engineering and Design

摘　　要：过去的定位研究大多是在结构化环境下进行研究的,但在野外环境下由于光照变化大、特征提取困难和颠簸道路多导致定位困难。提出一个由单目相机和惯性测量单元组成的可视化视觉惯导里程计,前端通过融合基于泊松方程预处理的视觉信息和惯性测量单元信息,后端采用非线性优化并结合鲁棒核函数使系统能够在光照变化大和颠簸道路的野外环境下鲁棒运行。实验结果表明,算法在ROOAD野外环境数据集中的表现比最先进的视觉惯导方法精度更高,鲁棒性更强,在EuRoC结构化数据集中的表现也超越现阶段最先进的算法。Most of the past localization studies were conducted in structured environments,but localization is difficult in field environments due to large changes in illumination,difficulty in feature extraction and many bumpy roads.The visualization monocular visual inertial odometry consisting of a monocular camera and an inertial measurement unit was introduced,and the front end of the system could operate robustly in the environment of large changes in field light,low texture and bumpy roads by fusing visual information and inertial measurement unit information at the front end,and using nonlinear optimization at the back end.Results of experiments show that the performance of the proposed method is more accurate and robust than that of the state-of-the-art visual inertial methods in the ROOAD off-road environment dataset,and outperforms the state-of-the-art algorithms in the EuRoC structured dataset.

关 键 词：同时定位与建图 定位 单目相机 惯性测量单元 野外环境 自动驾驶 环境感知 

分 类 号：TP242.62[自动化与计算机技术—检测技术与自动化装置]