The Indoor Wireless Location Technology Research Based On WiFi

In recent years, the current situation perception calculation become a hotresearch object of society at home and abroad. Content of Situation is very rich,and the most important one is the content of the position. How to calculate thephysical location is an important content of situational awareness to calculate.The most successful Research results are global positioning system and cellularnetworks. They are able to meet people in the outdoor environment of wirelesslocation requirements, But in the indoor environment, there is no more perfectwireless positioning solution. This article mainly discuss the indoor wirelesslocation technology based on WiFi, Researching based on fingerprint method ofaccess point selection and location estimation algorithm.First we introduce the main technology of indoor location and developmentsituation of indoor location. After this we expound the positioning methods,mainly locating methods can be divided into associated wi th distance and hasnothing to do with distance. In indoor wireless location based on WiFi, indoorenvironment is complex, the size of the location area is restricted, so geometricpositioning method based on distance is not suitable. The Fingerprint, that is,location fingerprinting method is much fit for indoor wireless locationenvironment compared with other methods. Its advantage is that positioningaccuracy is higher, and defect is that in offline stage it needs to collect a largenumber of fingerprint generation samples library, and workload is bigger.The main research content of this article is based on fingerprint method of APselection and location estimation algorithm. Now the most of AP selectionmethods work in the offline stage, they can reduce the amount of calculation well,but can not filter the APs affected by environmental performance variation. OurAP selection method work on the online phase. We introduce RANSACalgorithm used in image processing art to AP selection in the online stage forexternal detection. It can filter to remove the APs impacted by environmentalvariation, not only reduces the amount of calculation but also improves thepositioning accuracy. In the subsequent localization algorithm research, thetraditional bayesian algorithm and KNN algorithm are the most widely used twokinds of methods, but they both have their own defects. Aiming at thesedisadvantages, we improve the two kinds of algorithms. Based on traditionalBayesian algorithm, we adopt the concept of a regional division. Firstpreliminary positioning the backlog site area, further positioning point within thearea of the specific position. And on the regional location sum of probability is used to replace product of probability calculation. Classification based on thetraditional KNN algorithm is introduced into cluster and the cluster partition,allows a reference point to be assigned to multiple clusters, using differentfingerprint in different clusters. Finally we adopt a new method of dynamicunion combined with the above two kinds of improved algorithm. Based on theabove research, the average error of our positioning system is1.63meters, theminimum error is0.76meters.