Research On Fish-eye Image Aiding GPS/BDS Positioning By Mobile Phone In Urban Area

With the construction of GNSS system and the popularity of smart phones,the demand for location services in urban environment has soared,and the research of urban regional positioning based on low-cost terminals has become a hot topic.When we use phones to locate our location,there are two major problems to solve.One is the impact of complex environment on positioning performance,and the other is the data quality degradation caused by the smartphone antenna.Positioning in urban area is restricted by indirect signals(Non-Line-Of-Sight,NLOS)and multipath effect,and the signal quality obtained by low-cost equipment is also far from that of professional receivers.Considering the completion of Bei Dou Satellite Navigation System(BDS),the GPS/BDS positioning research based on smartphone with fish-eye lens is carried out.Three main research points are the environmental modeling assistance of image and map information,the detection and processing of abnormal signals,and the matching algorithm:(1)A method of fish-eye image information extraction based on double filter thresholding and pose correction is proposed.The useful information is extracted to improve positioning accuracy.With the camera,sensor data and 3D map provided by the smartphone,the environmental shielding situation around the user's location can be obtained.In order to meet the demand of real-time positioning and low power consumption,a dual filtering image segmentation scheme was proposed and the fisheye image was corrected based on pose information.The pixel classification accuracy of this method is 93.628%,and the average segmentation time is 3.08 seconds,which is nearly three times shorter than the K-means segmentation,and about 10% higher than the simple threshold segmentation accuracy.According to the principle of light propagation and 3D map data,the occlusion area is modeled,and the relationship among satellite data,image data and map data is established by using the skyline height angle of the blocked area.(2)This paper studies GPS and BDS signal quality and the suppression effect of weighted model on multipath effect in urban areas,then proposes a NLOS signal detection method integrating multiple data sources.For the problem of NLOS signals,the detection methods with multiple information sources are compared.The NLOS decision error rates using GNSS signal-to-noise ratio,3D map information and visual information are 15.25%,17.26% and 17.22% respectively.Considering the error caused by each data,the proposed comprehensive NLOS decision method reduces the error rate to 12.69%.The result shows that reducing NLOS signal weight can improve the positioning accuracy in densely built area,and the accuracy of east,north and up directions in the urban canyon is improved by 18.85%,22.81% and 14.30% respectively.(3)FE-SM(Fisheye Shadow Matching)algorithm is proposed.In view of the limitations of pseudorange positioning with smartphone in urban canyon,FE-SM algorithm uses the real-time visual information provided by fish-eye camera to improve the traditional shadow matching(SM)algorithm,such as optimizing the search area,determining and reweighting the street side score,and expanding the matching features.Compared with the initial positioning,the accuracy of Shadow Matching improves by46.70% on the along-street side and 63.53% on the cross-street side.FE-SM algorithm reduces the positioning error caused by the multiple high-score areas of traditional matching algorithm.While ensuring the accuracy of cross street positioning,it further improves the accuracy along the street by 47.63% and enhances the robustness of location in complex environment.