Research On Indoor Location Algorithm For Smart Parking System

With the improvement of people’s quality of life,cars have become a necessary way of transportation for people.At the same time,the number of cars has also increased in the city,which has put great pressure on the city’s transportation system and environment.In recent years,due to the shortage of parking space in the city,the situation of parking problem and random parking is serious.It seriously affects the normal traffic order.The reason why parking is difficult is not only the lack of parking spaces in the city but also the lack of parking space information.Due to the complexity of the parking lot structure,if the parking system lacks indoor positioning and navigation,people can not find parking space.Therefore,it is not enough to increase the number of parking spaces to solve the problem of difficult parking.It is also important to design a smart parking system to manage the parking spaces,private parking spaces and roadside parking spaces in urban parking lots.In addition,People can get parking information and book parking space online.Meanwhile,The parking system has the function of positioning and navigation,confirming information automatically and online payment.This paper analyzes the demand of parking system for parking problems and proposes a smart parking system based on Narrow Band Internet of Things.The smart parking system is designed and layered.It mainly includes parking lock,NB-IOT,data server and client application.In order to solve the problem of confirming information,Mobile phone and parking lock can automatically confirm each other’s information by bluetooth.In addition,The parking lock detects parking space status with low power consumption by multi-sensor.Based on the Bluetooth of the parking lock,the extended fingerprint data is studied.The first condition for fingerprint positioning is to build a complete and accurate fingerprint database offline.However,the indoor parking lot has a large area and it is difficult to construct a fingerprint database.To solve the problem,we expand the small Bluetooth fingerprint database by Generative Adversarial Networks and Pearson.It can not only reduce the time that it takes to collect fingerprint data,but also enhance Bluetooth fingerprint data information.Experiments show that,compared with the small Bluetooth fingerprint database,the accuracy is improved when we position by the expanded Bluetooth fingerprint database.Based on the expanded Bluetooth fingerprint database,a DNN-WKNN positioning algorithm is proposed.This algorithm combines Deep Neural Networks localization algorithm and Weighted K-Nearest Neighbor localization algorithm.Due to data noise,the positioning error of the single algorithm is large.However,These two algorithms compensate each other to reduce error.It has been proved by experiments that the DNNWKNN algorithm has a big advantage comparing the single DNN algorithm and WKNN algorithm in error mean and error variance.So the stability of positioning is improved.