Research On Indoor Localization Based On Regional Adaptive Fingerprint Analysis

With the emergence of problems such as parking space detection in intelligent traffic and personnel management and deployment in intelligent factory,it indicates that human beings have an increasingly strong demand for location-based services.Among them,obtaining the location of indoor objects is the core problem.Outdoor locations are positioned using Global Positioning System,but due to the complex indoor environment,it is no longer suitable for indoor positioning.At present,the research on indoor positioning mainly works from two aspects.One is to develop better positioning technology and positioning system,and the other is to optimize the positioning algorithm.The latter is the main factor for achieving highprecision positioning.The positioning algorithm is researched and improved,and the improvement effect is verified through experiments.First,the fingerprinting positioning theory is introduced.For k-means clustering in the offline stage,improper selection of the initial clustering center causes the problem of poor clustering effect.In this paper,a dual clustering method is used.for the first time based on the reference node coordinates clustering,select the reference point closest to the clustering center after the first clustering,and use the corresponding signal intensity fingerprint as the initial center of the second clustering.This method can effectively select k position coordinates that are far away.The fingerprint strength is optimized to optimize the clustering results;and the clustering process uses a new cost function to consider the similarity of fingerprints in the same region and between different regions.Experiments show that the algorithm in this paper can guarantee that the positioning result of nearly 80% of the points to be located is better than 1.6m,while the other two algorithms only have 34% and 13% positioning accuracy within 1.6m,and the positioning timeliness does not change significantly.Secondly,for the weighted K nearest neighbor method,different K values mean different numbers of fingerprints,which may lead to different positioning accuracy.To solve the problem of using a fixed K value of the weighted K nearest neighbor method when the fingerprint positioning method is used for online matching,a method based on the weighted K nearest neighbor method is proposed.The dynamic weight sum method adaptively adjusts K in different places and different networks to dynamically achieve the best accuracy in different places and different networks.Experiments show that the positioning results of the proposed method are closer to the optimal positioning results of the points to be located.Respectively,compared with the nearest neighbor method and the weighted K nearest neighbor method,the average positioning accuracy of the proposed method is improved by 12.8% and 8.9%.Finally,through the content of the second and third chapters,we can know that the accuracy of the fingerprint database is related to the positioning results.In order to establish an accurate fingerprint database,the data needs to be processed.In this paper,the z probability method is used to process the data.Experiments show that this method can effectively eliminate outliers with large fluctuations,and use data processed by different methods to achieve positioning.The positioning error corresponding to the z-probability method converges faster and the positioning effect is better.For the establishment of offline fingerprint database,this paper builds a set of positioning system based on ZigBee.The signal strength value at the anchor node received at the reference point and its own coordinate value form a fingerprint,and multiple fingerprints form a fingerprint library;and the characteristics of the collected data are analyzed.