基于视觉辅助优化的惯性可通行地图构建与行人定位方法

Inertial-based passable map construction and pedestrian localization method grounded in vision-assisted optimization

针对以惯性传感器为核心的室内行人导航系统存在长时间高精准定位问题,在基于历史信息约束的行人惯性定位方法基础上,提出了一种基于视觉辅助优化的惯性可通行地图构建与行人定位方法。以惯性传感器为核心,引入低频视觉信息优化行人轨迹,在滤波框架下加入优化线程。在粒子滤波的基础上,提取图像的VLAD特征作为地标特征,结合位置信息与地标信息进行地标特征匹配,根据匹配结果采用光束平差法基于拓扑结构优化行人轨迹,最终通过粒子重采样实现位置误差修正。在总长度1450 m的楼宇行走实验中,与基于历史信息约束的行人惯性定位方法相比,所提方法的平均定位误差为1.50 m,定位精度提升约40%。

To address the long time and high precision localization problem of indoor pedestrian navigation systems based on inertial sensors,an inertial-based passable map construction and pedestrian localization method grounded in vision-assisted optimization is proposed based on the pedestrian inertial localization method with history information constraint.With the inertial sensors as the core,low-frequency visual information is introduced to optimize pedestrian trajectories,and the optimization threads are added under the filtering framework.Based on the particle filter,VLAD features of the image are extracted as landmark features,and landmark feature matching is performed by combining position information with landmark information.Based on the matching results,the pedestrian trajectory is optimized using the bundle adjustment method according to the matching result.Finally,the position error is corrected by particle resampling.In the walking experiments of a building with a total length of 1450 m,the average positioning error of the proposed method is 1.50 m,and the positioning accuracy is improved by about 40%compared with the pedestrian inertial positioning method based on historical information constraints.