点线联合的优化视觉惯性里程计  

Point-and-line joint optimization visual inertial odometer

作　　者：危双丰 师现杰 刘振彬 肖斌 WEI Shuangfeng;SHI Xianjie;LIU Zhenbin;XIAO Bin(School of Geomatics and Urban Spatial Informatics,Beijing University of Civil Engineering and Architecture,Beijing 102616,China;Engineering Research Center of Representaive Building&Architectural Heritage Database for Ministry of Education,Beijing 102616,China;Key Laboratory of Urban Spatial Information,Ministry of Natural Resources,Beijing 102616,China;Beijing Key Laboratory for Architectural Heritage Fine Reconstruction&Health Monitoring,Beijing 102616,China;Jiangxi Nuclear Industry Surveying and Mapping Institute,Nanchang 330038,China)

机构地区：[1]北京建筑大学测绘与城市空间信息学院,北京102616 [2]代表性建筑与古建筑数据库教育部工程中心,北京102616 [3]自然资源部城市空间信息重点实验室,北京102616 [4]建筑遗产精细重构与健康监测北京市重点实验室,北京102616 [5]江西核工业测绘院,南昌330038

出　　处：《测绘科学》2021年第4期20-27,36,共9页Science of Surveying and Mapping

基　　金：国家自然科学基金项目(41601409);北京市自然科学基金资助项目(8172016);北京建筑大学市属高校基本科研业务费专项资金资助项目(X18229);北京建筑大学研究生创新项目(PG2018066,PG2019065,PG2019061)。

摘　　要：为了减弱视觉同时定位与地图构建(SLAM)易受光照、纹理等条件的影响,在非线性化单目SLAM研究基础上,该文提出了一种加速度计bias估计优化初始化与点线特征结合的优化视觉里程计,使得在光线较弱的情况下也有较好的位姿估计效果,更优秀的初始化结果使得整个系统更加鲁棒、精度更高,且为了减少因为线特征的加入而增加的计算量,提出一种新的数据选择策略。通过和其他优秀算法(如PL-VIO、仅特征点方案)对比及真实场景实验的结果分析可知,该文提出的点线联合的优化视觉惯性里程计不仅能够减少定位误差,而且在光照条件较弱的环境中有较高的精度,既保证了系统的实时性,又提高了系统的鲁棒性。In order to reduce the visual simultaneous localization and mapping(SLAM)’s susceptibility to lighting, texture and other conditions, an optimized visual odometer combined with optimized initialization of the bias estimation of the accelerometer and point-line features is proposed which makes better pose estimation effect even in the weaker lighting conditions,based on the study of nonlinear monocular SLAM.The better initialization results make the whole system more robust and more accurate.In order to reduce the amount of calculation added due to the addition of line features,a new data selection strategy is proposed.Through comparison with other excellent algorithms and the analysis of the experimental results of real scene experiments,it can be seen that the point-and-line joint optimization visual inertial odometer proposed in this paper can not only reduce positioning errors,but also have higher accuracy in environments with weak lighting conditions,which guarantees the real-time performance of the system,and the robustness of the system is also improved.

关 键 词：同时定位与地图构建(SLAM) 线特征 非线性优化 初始化 鲁棒性 

分 类 号：P228.4[天文地球—大地测量学与测量工程]