基于条件随机场模型的室内地图匹配算法  

Indoor map matching algorithm based on conditional random field model

作　　者：郭红雨 任明荣 王普[1,2] Guo Hongyu;Ren Mingrong;Wang Pu(College of Automation,Faculty of Information Technology,Beijing University of Technology,Beijing 100124,China;The Ministry of Education P.R.C,Engineering Research Center of Digital Community,Beijing University of Technology,Beifing 100124,China;Beifing Key Laboratory of Computational Intelligence and Intelligent System,Belting University of Technology,Beijing 100124,China)

机构地区：[1]北京工业大学信息学部自动化学院,北京100124 [2]北京工业大学数字社区教育部工程研究中心,北京100124 [3]北京工业大学计算智能与智能系统北京市重点实验室,北京100124

出　　处：《仪器仪表学报》2018年第5期107-114,共8页Chinese Journal of Scientific Instrument

基　　金：北京市教委项目(KM201610005006)资助

摘　　要：基于微机械系统（MEMS）技术的惯性导航系统（INS）,在室内定位领域受到了广泛的关注。为解决其定位精度随时间发散的问题,学者们提出融合室内地图的解决方法。但是目前的室内地图匹配算法存在匹配正确率不高、计算量大等问题。为了提高匹配的正确率,研究了一种基于条件随机场模型的地图匹配算法。算法采用闭环设计,将行人的最优匹配点作为个人导航系统的反馈量,对惯性导航系统输出的位置进行修正。位置修正提高了惯性导航输出位置信息的精度,因此将观测点坐标作为条件随机场模型的一个特征量。坐标点的提取是以行人行走的固定长度为依据,相对于已有文献中基于步长为坐标点的提取方式,该模型结构减小了每次状态点选取的个数,从而减少了算法的计算量。多次实测结果表明,该算法提高了地图匹配的正确率。The inertial navigation system based on micro-electro-mechanical-system（MEMS） has received extensive concern in the field of indoor positioning. In order to solve the problem of positioning precision divergence with time,scholars proposed the solutions of fusing indoor map. However,in current indoor map matching algorithms there exist the problems of low correct map-matching rate,large computation burden and etc. In order to improve the correct map-matching rate,this paper presents a novel map-matching algorithm based on conditional random field model. The algorithm adopts closed loop design,and takes the optimal matching point of the pedestrian as the feedback quantity of the personal navigation system to correct the position outputted from the inertial navigation system. The position correction improves the accuracy of the inertial navigation system output position information; then,the observation point coordinates are considered as a feature quantity of the conditional random field（CRF） model. The extraction of the coordinate point is based on the fixed length that the pedestrian walks. This model structure reduces the number of the state point extraction compared with the extraction method taking step length as the coordinate point in some existing literatures,thereby the computation burden of the algorithm is reduced. Multiple experiment results show that the proposed algorithm improves the correct rate of map matching.

关 键 词：微机械系统 个人导航 条件随机场 地图匹配 

分 类 号：TH702[机械工程—仪器科学与技术] U666.1[机械工程—精密仪器及机械]