Research On Geomagnetism/PDR Indoor Fusion Localization Technology Based On Particle Filter

As one of the hottest research fields at present,indoor positioning has attracted many enterprises,universities and research institutions at home and abroad to invest in it.As one of the indoor positioning techniques,geomagnetic positioning is widely concerned due to its features such as wide global coverage,no need to set up infrastructure,low hardware cost,high positioning accuracy,strong stability and no accumulation of positioning errors over time.However,due to the limited features available for geomagnetic positioning,single positioning method is prone to mismatching,which leads to serious rebound problem of the location.In addition,geomagnetic positioning is easy to be affected by the change of pedestrian motion state,which will also have a certain impact on the accuracy of geomagnetic positioning.In view of the above problems,this thesis implements a geomagnetic indoor positioning system based on smartphone platform.This system uses particle filter algorithm to integrate pedestrian dead reckoning and geomagnetic positioning technology and realizes real-time visual display of positioning results in smarphone.In this thesis,the algorithm of pedestrian position calculation and geomagnetic positioning is optimized and improved to improve the overall positioning effect of the system.The main research contents and achievements of this thesis include:1.Improve the accuracy of pedestrian dead reckoning.In order to reduce the offset accumulated over time by PDR positioning technology,we optimized the key technology in PDR positioning.For step recognition,the finite state machine method with high recognition rate and low algorithm complexity is adopted as the step counting algorithm in this thesis,and higher detection accuracy is obtained under different motion states and different mobile phone holding modes.In order to obtain accurate and reliable step model,we used a large number of training samples to fit the model parameters with an average accuracy of 0.02 m.In the aspect of azimuth estimation,this thesis takes human motion characteristics into consideration and uses the two-step azimuth mean of electronic compass as the azimuth at the current moment,which effectively solves the problem of periodic fluctuation of azimuth and the accuracy of azimuth estimation is better than complementary filtering.2.Improve the accuracy and reliability of geomagnetic positioning.To solve the problem that geomagnetic fingerprint sequence matching is easily affected by the change of pedestrian motion state,we propose a sequence matching algorithm thattakes pedestrian speed into consideration.This algorithm combines the step detection and step length estimation in PDR positioning to resample the real-time fingerprint sequence to adapt to the resolution of geomagnetic map.Practice has proved that the improved sequence matching algorithm avoids duplicate establishment of fingerprint database and has high practical value in real-time location.Amount,based on the geomagnetic single positioning way is easy to produce false matching problem,we put forward a kind of geomagnetic features more collaborative localization algorithm,the algorithm gives up the decision power of single positioning method to the positioning result,use two localization methods to judge the credibility of location,can effectively improve the positioning accuracy,reduce the rebound problem in the process of magnetic positioning.3.Improve the robustness of the system.We conduct particle filter fusion for geomagnetic positioning technology and PDR positioning technology to realize the complementary advantages of the two positioning methods.Since the key algorithm in the two positioning technologies is optimized and improved in this thesis,it can provide the high-quality state quantity and observation quantity,and effectively ensures the improvement of the overall positioning accuracy of the system.The integrated pedestrian track basically eliminates the drift of PDR,and solves the problem of geomagnetic positioning jumping back and information loss.The average positioning accuracy reaches 1.38 m and the standard deviation is 0.96 m.The accuracy and stability of the system have been greatly improved.4.Realize actual availability of the system.In order to apply the fusion localization algorithm to the real-time localization,we divide the hierarchical structure and the localization module of geomagnetic localization in detail,and analyze the data interaction relationship between each module.On this basis,a set of geomagnetic positioning system for real-time positioning and map visualization display is implemented on smartphone by using Eclipse development software.Through the experimental evaluation of the actual indoor scene,it is found that the system can obtain real-time positioning results with high accuracy and stability.