The Research Of 3D Autonomous Indoor Location System Based On MEMS Sensor Of Intelligent Terminal

In recent years,indoor positioning and navigation have been extensively studied in the academic and industrial fields and have developed rapidly.However,the current indoor positioning technology has obvious defects: a large number of auxiliary equipments need to be arranged in advance,the cost is high and the practicability is reduced;or the pace model fixed and the gait model is single in a pedestrian dead reckoning(PDR)algorithm,two-dimensional display of the track,the positioning accuracy can not reach the effect of actual use.Currently,smart terminal devices are represented by mobile phones or tablet devices equipped with an Android or iOS operating system,and integrate accelerometers,gyroscopes,magnetometers,barometers and other micro electro mechanical system(MEMS)sensors which have a small size,low power consumption and low cost.MEMS sensors,powerful intelligent CPU,and high speed large capacity memory,which provide a mature software and hardware platform for indoor navigation and positioning technology.This thesis aims at the uncertainty of using smart devices for a pedestrian,researches the principle and system architecture of pedestrian positioning system,and designs an indoor positioning algorithm with multiple pedestrian states and equipment multi-poses,thus effectively solving the problem of pedestrian navigation and positioning in daily indoor scenes.Firstly,the error calibration model of accelerometer and magnetometer is established for low-cost MEMS sensors on smart phones.Through simulation analysis and experimental verification,the error calibration model effectively solves the impact of low-cost sensors errors on the pedestrian positioning system,and provides a basis for subsequent positioning algorithms.Secondly,a machine learning algorithm based on support vector machine(SVM)is proposed to effectively classify the sensor output data in multiple modes of a pedestrian.It can identify four movement states of a pedestrian and four poses of mobile phones and the experiment proves that the recognition rate is over 85%.At the same time,based on the correct discrimination of multiple modes,a full-attitude positioning algorithm is designed,including the full-attitude step detection algorithm,attitude fusion algorithm and heading calibration algorithm.It effectively solves the problems of missing steps and heading mutations in different mobile phone attitude switching processes.And aiming at the problem that the cumulative error divergence of inertial sensors during walking for a long time,a virtual landmark matching and compensation algorithm based on indoor map is proposed.The cumulative error is eliminated by resetting the starting point.Then this thesis combines the relative height of a pedestrian calculated by the barometer output data and motion status recognition,to determine the floor of a pedestrian.Finally,the algorithm is verified through Android platform,and the algorithm is placed in the system running layer in the form of a shared library through JNI(Java Native Interface,JNI)technology,which effectively guarantees the operating efficiency of the algorithm.The test results show that the system can provide accurate and effective positioning information.Eventually,the indoor positioning error of multiple scenes is within 3% and the compliance rate is over 85%,which proves the excellent performance of the algorithm.