面向室内弱光环境的视觉与惯导融合算法研究  

Research on fusion algorithm of vision and inertial navigation for indoor low light environment

作　　者：王孟奇 连增增[1] 田亚林 王鹏辉 WANG Mengqi;LIAN Zengzeng;TIAN Yalin;WANG Penghui(School of Surveying and Land Information Engineering,Henan Polytechnic University,Jiaozuo,Henan 454003,China)

机构地区：[1]河南理工大学测绘与国土信息学院,河南焦作454003

出　　处：《导航定位与授时》2024年第5期25-35,共11页Navigation Positioning and Timing

基　　金：国家自然科学基金(42374029);河南省自然科学基金(202300410180);河南省高校基本科研业务费专项资金(NSFRF230405);河南理工大学2017年度博士基金资助项目(B2017-10);河南理工大学青年骨干教师资助计划(2022XQG-08)。

摘　　要：针对视觉惯性里程计算法在低光照或黑暗等复杂条件下定位精度差的问题,提出了一种基于多状态约束卡尔曼滤波(MSCKF)算法的适应弱光环境的视觉惯性里程计算法。结合离散余弦变换(DCT)同态滤波和限制对比度的自适应图像均衡化处理图像,然后将增强后的图像应用于MSCKF中,计算出精确的初始位姿估计。基于多状态观测约束策略计算特征重投影误差,以更新系统状态。该算法在公开数据集EuRoC上进行了测试。结果表明,不管在正常光照还是弱光照的场景下,该算法相较于原始算法精度均有所提升,最大均方根误差从1.778 m降低到0.249 m,平均均方根误差降低57.7%。Aiming at the problem of poor positioning accuracy of visual inertial odometry algorithm under complex conditions such as low light or darkness,a visual inertial odometry algorithm based on multi-state constraint Kalman filter(MSCKF)is proposed to adapt to low light environment.The image is processed by combining discrete cosine transform(DCT)homomorphic filtering and adaptive image equalization with limited contrast,and then the enhanced image is applied to MSCKF to compute the accurate initial pose estimation.Finally,the feature reprojection error is calculated based on the multi-state observation constraint strategy to update the system state.The algorithm is tested on the public dataset EuRoC.The results show that the accuracy of the algorithm is improved compared to the original algorithm in both normal illumination and low illumination scenarios.The maximum root mean square error is reduced from 1.778 m to 0.249 m,and the average root mean square error is reduced by 57.7%.

关 键 词：视觉惯性里程计算法 同态滤波 限制对比度的自适应直方图均衡化 弱光环境 

分 类 号：V249.31[航空宇航科学与技术—飞行器设计]