The Research And Implementation Of Radio Frequency Fingerprint Effectiveness Technology Based On Crowdsourcing

In recent years,indoor positioning technology continues to mature,many positioning technologies have entered the commercial stage.With the continuous development of mobile Internet,plenty of wireless access points have been deployed in various department stores and parking lots.Therefore,the wireless positioning algorithms will be able to achieve positioning without eploying equipment additionally and have been widely used.Indoor positioning experts and professional devices are necessary for traditional wireless localization algorithms to sample fingerprints and construct radio maps.These methods not only are time-consuming and laborious,but also seriously impedes the commercial promotion of wireless localization algorithms.Hence,more and more scholars have proposed wireless localization algorithms based on crowdsourcing.However,the effectiveness of collected fringerprints can't be guaranteed because of the instability of the signal sources,the non-professional ordinary users and the intrusion of malicious users.The paper studies the current status of research about the application of radio frequency fingerprints in indoor localization.From the perspective of the fingerprints' effectiveness,this paper designs and achieves ODBS(Outlier Detection Based on Similarity)algorithm on the basis of the fingerprints similarity.The algorithm leverage the plenty of sensors(acceleration sensor,magnetic sensor,Wi-Fi sensor,etc.)embedded in smartphones to sample fingerprints that exist widely in the physical world,and uses the AFP(AFfinity Propagation clustering algorithm)and DTW(Dynamic Time Warping)to classify fingerprints,and the DTW algorithm is utilized for fingerprint fusion.WeDiff(Wireless fingerprint Effectiveness Detection based on RSS difference)and WeFree(Wireless fingerprint Effectiveness Detection based on Freeloc)are designed to detect the fingerprints' effectiveness.KNN-DTW(K Nearest Neighbor Dynamic Time Wraping)is designed to detect the outliers.The algorithms as noted above all are based on the atomic trajectory.The atomic trajectory classification algorithm based on AFP and DTW calculates the similarities of atomic trajectories by seeking the intersection while eliminating differences,and thus obviously improves the accuracy of classification.As to fuze the wireless radio fingerprints,the prematch between three-dimensional geomagnetic fingerprints is finished firstly,and then fuze the wireless radio fingerprints according to the prematching result,thus spatial characterization ability of the fused atomic trajectory is enhanced greatly.WeDiff and WeFree can detect the failed atomic trajectory effectively,KNN-DTW can eliminate the possible outliers in the atomic trajectory.To test the performance of the ODBS,the paper designs and implements the WiMap(Wireless Map)algorithm to build a wireless fingerprint map based on crowdsourcing.The WiMap solves the difficulty of building a traditional wireless fingerprint map.The paper designs and implements the ODBS system and WiMap system,and evaluates the performance of each algorithm respectively.The experimental results show the algorithms proposed in this paper can achieve better performance in detecting the outliers and effectivenss of fingerprints,and thus improve the positioning performance.WiMap has achieved the same positioning accuracy as the traditional building algorithm while eliminating the time-consuming and laborious work in the training phase.WiMap will play a positive role in the commercial application of wireless indoor positioning technology.