为了降低行人航位推算(Pedestrian dead reckoning,PDR)算法在进行井下人员定位时产生的累积误差,提出了一种基于PDR算法与伪平面技术的井下人员定位方法.首先,采用惯性导航传感器获取井下人员的步态信息,通过线性步长估计模型和四元数法实现步长估计和方向估计,利用PDR算法推算人员的位置;其次,使用井下人员活动区域以及预设的标记点构建伪平面,并将井下人员位置映射到伪平面坐标上,为降低PDR算法的累积误差做准备;最后,采用SVM进行井下人员活动检测,通过转弯活动判断其是否处于特殊标记点,将PDR解算的位置与伪平面内已知转弯位置标记点进行相关性分析,完成伪平面信息与工人位置的匹配,校准并更新PDR位置,降低累积误差.结果表明:井下工人在完成单个转弯活动过程中,传统PDR算法解算位置平均误差为0.98 m,而进行伪平面修正后平均误差降低到0.31 m;在完成区域性多活动过程中,采用伪平面技术修正后的PDR平均定位误差从1.08 m降低到0.38 m.因此,所提出的井下人员定位方法有效提高了 PDR算法的定位精度.
Abstract
Cumulative error occurs when pedestrian dead reckoning(PDR)algorithm is adopted to posi-tion underground personnel.To solve this problem,a personnel location method based on PDR algo-rithm and pseudo-plane technology is proposed in this paper.Firstly,the gait information of under-ground personnel is obtained by inertial navigation sensor,the step size estimation and direction estima-tion are achieved by linear step size estimation model and quaternion method,and the position of per-sonnel is determined by PDR algorithm.Secondly,the pseudo-plane is constructed by using the activity area of underground personnel and the preset marking points,and the location of underground personnel is mapped to the pseudo-plane coordinates for reducing the cumulative error of PDR algorithm.Finally,SVM is used to detect underground personnel activities,and the turning activity is used to determine whether they are at special marks.A correlation analysis is performed between the position calculated by PDR and the known turning position marks in the pseudo-plane,so as to match the pseudo-plane in-formation with the workers'positions,calibrate and update the PDR position,and reduce the accumula-ted errors.The experimental results show that the average position error of the traditional PDR algorithm is 0.98 m,while the average error is reduced to 0.31 m after the pseudo-plane correction.In the process of regional multi-activity,the average positioning error of PDR after calibration with pseudo-plane technology is reduced from 1.08 m to 0.38 m.Therefore,the positioning method proposed in this paper could improve the positioning accuracy of PDR algorithm effectively.
维普
万方数据