双目视觉辅助PDR的组合导航定位方法

Combination Navigation and Positioning Method for Binocular Vision Assisted PDR

为提高室内定位系统精度和跟踪性能以及适应复杂环境,将行人航迹推算(Pedestrian Dead Reckoning,PDR)与双目视觉组合,提出一种双目视觉辅助PDR的组合导航定位方法.该方法通过选取或布置地标建立了地标位置数据表;基于轻量化目标检测实现了对地标实时双目测距,保证定位的实时性;利用PDR位置信息得到检出地标类别对应坐标,基于因子图的协同定位和误差估计算法将双目视觉与PDR有效融合,提高了定位精度并抑制PDR累计误差,同时对PDR中航向和单参数模型中单位转换常数进行误差补偿,提高PDR定位精度.实验结果表明,在地标纹理清晰且分布合理情况下,该方法能有效解决室内复杂环境下单一PDR累积误差问题,此外,对航向和单位转换常数实时补偿可提高组合定位系统的定位精度和稳定性.

In order to improve the accuracy and tracking-performance of indoor positioning system and adapt to com⁃plex environment,a new integrated navigation and positioning method based on binocular vision assisted pedestrian dead reckoning(PDR)was proposed by combining PDR with binocular vision.In this method,the landmark position data-table was established by selecting or arranging landmarks.Based on lightweight target detection,the real-time binocular ranging of landmarks was realized.Using PDR position information,the coordinates corresponding to the detected landmark catego⁃ry were obtained.The cooperative positioning and error estimation algorithm based on factor graph effectively integrated binocular vision with PDR to improve the positioning accuracy and suppress the PDR cumulative error.Simultaneously,the error compensation was carried out for unit conversion constants in azimuth and single-parameter model,which improved the PDR positioning accuracy.The results show that this method can effectively solve the cumulative error of single PDR in a complex indoor environment with clear and well-distributed landmarks,and the real-time compensation of heading and unit conversion constants can improve the accuracy and stability of the integrated positioning system.