ArUco-SLAM:基于ArUco二维码阵列的单目实时建图定位系统

ArUco-SLAM:a Monocular SLAM System Based on ArUco Landmark Array

高精度自动建图算法是ArUco二维码阵列定位系统应用的前提和核心,现有在线建图算法累计误差大且易受误检测影响,而人工标定和离线建图的方式则效率低、耗时长.为解决上述问题,本文提出了一种ArUco-SLAM单目实时建图定位算法,其采用前后端并行建图、联合优化的框架,前端以坐标系变换封闭原理对地图进行快速更新和局部修正;后端以非线性优化算法最小化二维码顶点总体重投影误差,实现对关键帧序列囊括地图的全局优化;最终,基于联合优化算法实现对前后端地图的匹配,消除前端地图的累计误差.设计实验对所提算法进行了验证,实验结果证明其相比现有建图系统有着累计误差小和建图效率高的优点,具有建立大规模二维码阵列厘米级精度地图的能力.

Accurate mapping algorithm is the core of ArUco landmark array localization.Most of the existing mapping algorithms based on offline processing or manual calibration behave inefficiently and time-consuming.In addition,the existing online mapping algorithms can though guarantee real-time performance,but show low accuracy,large cumulative error and sensitive to false detection.In order to overcome those shortcomings,a monocular ArUco-SLAM algorithm was proposed based on parallel mapping and synchronization optimization in this paper.Firstly,the closure principle of coordinate transformation was used to build and correct the global map online.And then,a linearity optimization algorithm was used to minimize the projection error in the corners and to achieve global optimization for key frame of the map.A synchronization optimization algorithm was used to match the online map and optimized map,and to reduce the cumulative error of the global map.Finally,several contrast experiments with other mapping systems were carried out.The experimental results show that the proposed system can not only establish high-precision map,but also realize online mapping and real-time camera localization.