A Landmark Aided PDR Indoor Positioning Method Extended By The Corner Recognizer

In recent years,pedestrian dead reckoning(PDR),Wi-Fi and a variety of fusion indoor positioning technologies based on mobile smart devices have been developed greatly.At the same time,the existing methods also have disadvantages that can't be ignored,such as larger deviation of positioning results and higher requirements on the positioning environment.For this issue,an indoor positioning method combining map information and classification recognition algorithm based on PDR is proposed in this thesis,which is called landmark aided PDR indoor positioning method extended by the corner recognizer(C-LaPDR).First of all,the PDR algorithm is used as the underlying localization algorithm of the system.Through the experiment,a real-time step estimation model based on step frequency and acceleration peak is established.This model can adapt to the complex situation of different walking frequencies of different pedestrians or the same pedestrian.To a certain extent,the versatility of the model is improved.Next,for the positioning environment,the corresponding landmark database is established,and different recognition models and position calibration algorithms are established according to different landmark types.The purpose is to correct the accumulated error in the underlying PDR positioning algorithm by means of physical condition constraints and fingerprint matching.Then,the diversity of mobile device poses carried by users is considered in this thesis,and a hierarchical corner recognition system composed of four classifiers is established.By comparing the classification effects of J48 decision tree,support vector machine and naive Bayesian classification algorithm,it is found that the correct rate of the J48 decision tree for the pose classification and corner recognition of the data set is more than 90%.This algorithm also takes relatively little time.Therefore,in this thesis,J48 decision tree is selected as classification algorithm.By training the pose classifier and the corner recognizer in three different poses,the hierarchical corner recognition system is established,which can reduce the mismatch of corner landmarks due to the occurrence of false corner during walking.Finally,the corridor of the 4th floor of J9 Building of Shandong University of Science and Technology is used as the experimental site,and the experiment is designed to compare the positioning effect of the C-LaPDR algorithm and the PDR algorithm.It can be seen from the experimental results that there is a drift section with large error in the PDR algorithm positioning trajectory.For the C-LaPDR algorithm,the drift error of the trajectory can be controlled within 0.5m for a period of time after the landmark matching.The C-LaPDR algorithm has a good correction effect on the trajectory drift through the process of landmark matching and position calibration.Overall,the average positioning accuracy of the system is improved to about 1.5m by the landmark aided PDR indoor positioning method extended by the corner recognizer.