Research On Location Fingerprint Location Algorithm Based On MapReduce

The global terrorist attacks happened frequently in public places.In all kinds of military action,street fighting can not be avoided.Thus,a small range of indoor positioning technology can help our army lock and hit the key targets better.The location fingerprint positioning technology overcomes the shortcomings of ranging and positioning with electromagnetic wave,which requires high clock synchronization and is influenced by the environment.The location fingerprint positioning technology provides a good technical support for accurate positioning requirements,especially for indoor positioning.In this thesis,based on the characteristics of the location fingerprint location and the extensive development of the current cloud computing,the thesis studies the different location fingerprint location algorithm and its realization in the cloud computing environment.The main contributions in this thesis include as follows:Based on the traditional location algorithm,a kind of NN-KNN algorithm is studied.It is the weighing process of NN(Nearest Neighbor)and KNN(K-Nearest Neighbor)algorithms.From the theoretical analysis,this thesis shows the advantages of the method compared with the traditional position fingerprint location algorithm.Experiments show that the algorithm has advantages in the accuracy of the positioning performance,positioning accuracy is improved significantly compared to the traditional positioning algorithm,the positioning probability of 2m positioning increased to 65%.In order to enhance the universality and flexibility of this algorithm,based on the NN-KNN algorithm,this thesis proposes an improved positioning algorithm KNN-Hybrid.The algorithm increases the amount of date with different value of k.the different k values corresponding to the positioning results weighted processing.This will not only get a better positioning performance,but also can adapt to different positioning environment.The experimental results show that the localization probability of the algorithm is 73%,compared to the NN-KNN algorithm,it is improved by about 8%in the 2m algorithm.A computational model suitable for KNN-Hybrid algorithm is studied.According to the characteristics of KNN-Hybrid algorithm,the use of cloud computing tools in the MapReduce computation model is used in a distributed framework technology advantage to process the algorithm,and add a layer of data filtering layer based on the traditional MapReduce computation model,It optimizes and improves the traditional MapReduce model,On the basis of the filtering redundant data for more suitable to KNN-Hybrid algorithm on the amount of data processing.The experimental results show that the improved MapReduce model can make the KNN-Hybrid algorithm to reduce the data processing capacity of 20%,and improve the positioning efficiency of the algorithm.