Crowdsensing Based Key Technologies Of High Accuracy Indoor Wifi Fingerprint Localization

As important individual attribute information of citizen and the key environment sensing target,indoor location information provides assistance to people’s work,entertainment and personal life.In resent years,benefited from the popularity of mobile intelligent terminals equipped with multi-sensor,large-scale commercial deployment and application of wireless networks and the rapid development of big data mining technology,Wi Fi based indoor positioning technology has stood out from many positioning technologies and received attention by users.However,the Wi Fi fingerprint based indoor location technology faces many challenges in the actual deployment and application,such as the high system deployment cost caused by labor intensive site survey,the static radio-map cannot adapt to the changes of dynamic environment and gradually lose its effectiveness,and the low positioning accuracy and instability of positioning results in different indoor functional areas.These problems greatly limit the application and popularization of fingerprint localization in indoor scenarios.The emergence of Crowd Sensing(CS)can alleviate the problem of high cost of site survey and fingerprint uploading untimely,and strengthen the accuracy,reliability and scalability of the positioning system.Although CS makes the site survey cheap and convenient,it will reduce the quality of fingerprint data.Location fingerprints collected in some areas could be sufficient and rich because volunteers pass by frequently.On the contrary,some special functional areas are lack or even missing of the relevant fingerprints due to the lack covered of volunteers path.In addition,radio-map updating requires volunteers to walk through a certain number of Reference Points(RP)to collect fingerprints in a short time,which cannot be satisfied by CS.And the unprofessional volunteers cannot collect and upload the unnoisy location fingerprints at all RPs.Finally,due to the indoor electromagnetic environment is complex,the distributions of fingerprint in different areas are uncertainty and different,which makes the positioning accuracy is limited and unstable.To solve the above problems,the main contents and contributions of the thesis are as follows.Firstly,To solve the problem of multi-dimensional location fingerprints incompletion in CS network,we investigate CS working mechanism and propose a tensor completion and sparse representation based fingerprint automatic generation algorithm.Begin with the analysis of the quality problems of fingerprint data in CS network,we point out the fingerprint provided by volunteers exist in the form of high-dimensional and incomplete.In most cases,Access Points’(AP)location are fixed in indoor environment,which generated a correlation naturally between the originally independent and uncorrelated fingerprints collected at each RP.Although the fingerprints at different RPs will change along with the environment,the volunteer action mode and indoor space structure will not change.Furthermore,due to the complex indoor scenario and the limited sensitivity of mobile devices,fingerprint data generally presents a certain sparsity.The algorithm we proposed can generate fingerprints by using the correlation between tensor location fingerprints in CS network,so as to provide high-quality fingerprint data for location services.Extensive experimental results show that the location fingerprints generated by proposed algorithm can improve the positioning accuracy in CS network.Secondly,to solve the decreased effectiveness of radio-map,we propose a matrix low-rank and sparsity and dictionary learning based radio-map updating algorithm.Begin with the analysis of the fingerprint sensitivity of indoor environment,we point out this phenomenon will ultimately reduce the positioning accuracy.Then,we find the correlation between fingerprints is little impacted by the changing environment.This kind of correlation between fingerprints is always expressed as the low-rank property of the fingerprint matrix.Meanwhile,the complex indoor environment will cause numerous fingerprint elements missing which expresses as a certain sparsity in the radio-map.Based on the theoretical analysis of the radio-map matrix,the radio-map updating algorithm we proposed can update the radio-map timely and automatically by using the low-rank property and sparse property.Experimental results show that the proposed algorithm can accurately update the fingerprint map and remove the noise.The positioning accuracy can be greatly improved by using the updated radio-map.Thirdly,to solve the problem of low positioning accuracy and poor stability of positioning results,we propose a perturbation based AP selection algorithm and a distance metric learning algorithm.Based on the analysis of the collected fingerprints,we point out the difference in AP discrimination capability,the difference in AP signal stability and AP redundancy are the main reasons causing low positioning accuracy.To solve the problem,we propose a perturbation based AP selection algorithm.The algorithm removes redundant AP by judging the influence of disturbance on the solution of fingerprint linear equation.And it can effectively improve the accuracy and stability of fingerprint localization and reduce the computational overhead of positioning.Furthermore,we propose a fingerprint distance metric learning based fingerprint positioning algorithm.It can extract the indoor environment information to product the RSS similarity metric matrix for the improvement of positioning accuracy.Experimental results show that the positioning accuracy and accuracy stability can be further improved by reasonably applying the proposed algorithms to the fingerprint localization.