Research On WiFi Indoor Location Algorithm Based On KPCA And PDR

Thanks to the rapid development of mobile Internet and the popularization of smartphones,Location Based Service(LBS)has been widely used in many fields.Although the application of GNSS in the field of outdoor positioning has been mature,the indoor positioning is limited by the conditions of time,precision and complex environment,and there are still some shortcomings,such as imperfect algorithm,irregular system and weak universality,which makes the indoor positioning algorithm become one of the research hotspots in the field of surveying and GIS.In the algorithm researches,sensor-based Pedestrian Dead Reckoning(PDR)has the advantages of high precision,simple deployment and easy analysis,and is widely used;Based on kernel principal component analysis,fingerprint dataset is mapped,the nonlinear relationship of the dataset in the higher dimension is refined,the storage capacity of the data is changed and the matching speed of on-line data acquisition is improved.In this paper,a WiFi indoor positioning algorithm based on KPCA and PDR is proposed,including the following 4 points:(1)This paper takes the smartphone as the research platform,collects the fingerprint information of the reference point location around its internal integrated WiFi module,obtains the fingerprint intensity value,then filters the dataset in real time to eliminate the noise error,and also makes the optimal estimation of the system state.(2)using the sensor of the smartphone,and the peak detection method based on angular velocity,the step numbers of pedestrian are detected,the frequency cycle of the pedestrian's gait under walking condition is analyzed;the pedestrian step length information is detected by using the nonlinear function model of step frequency and height,and the calculation formula of pedestrian step length under specific circumstances is obtained;based on the direction rotation of the gyroscope of the smartphone,the change of the heading angle of the pedestrian is collected as it moves around.(3)The main component analysis method based on Gaussian kernel function maps and transforms the offline multidimensional fingerprint data,refines the main feature attribute between the fingerprint intensity under the condition of reducing the dimension,reduces the data storage capacity,and also improves the speed and precision of the online stage matching.(4)Using the fusion location model of WiFi in KPCA and PDR and based on the system development environment of Android and the editing and processing of Java language,the overall architecture and simulation of the positioning system are proposed and constructed in this paper,and the simulation experiments are carried out in the experimental area.The data collected by the experiments are detected and analyzed,achieving the expected positioning effect.