Research And Application Of Bluetooth Indoor Positioning Technology Based On Deep Learning

In recent years,with the advancement of science and technology and the improvement of people's living standards,people's demand for the accuracy of indoor positioning services has been greatly improved.How to accurately obtain the user's precise indoor location has become a hot issue in indoor location-based services,which has given rise to extensive discussions and researches.At present,the mainstream indoor positioning technology can be basically divided into Wi Fi positioning,Bluetooth positioning,Zig Bee,RFID positioning,UWB positioning,geomagnetic positioning,inertial navigation positioning,ultrasonic positioning,etc.Among them,Wi Fi positioning technology and Bluetooth positioning technology are widely used in indoor positioning technology,with high integration and moderate cost.It is because of the similar positioning accuracy,cost,and coverage,that the two can often replace each other.But since the birth of Bluetooth technology 5.0standard,it is the lower power consumption,higher transmission speed,longer transmission distance and larger transmission capacity that have improved the accuracy of Bluetooth positioning,the bilateral information transmission,meaning that the transformation brings much more potential possibilities for the realization of wide-area Internet of Things technology,which leads to the growing favour among the academics.This paper studies the Bluetooth indoor positioning algorithm based on deep learning,establishes a RSSI-Distance Relation Positioning Model Based on Keras(RDRPK),and apply it to the reverse car-finding system for verification.First of all,the experiment of site equipment has been carried out as well as the positioning data has been collected in the basement,the initial data of the Bluetooth signal strength is closely combined through a series of data-pretreatment.Secondly,the RDRPK model choose the Keras neural network framework based on the analysis of the deep learning network construction framework.When the number of the framework-layers,function configuration and calculation times of the neural network has been determined,the high-accuracy RDRPK model has been proposed by taking the collected data as the learning sample to get the RDRPK model.At the last step,the reverse car-finding system is used to conduct a real-scene positioning experiment on the target user.After analyzing the error of positioning data and comparing the positioning accuracy of the front and rear reverse car-finding system using the RDRPK model,the paper draw a conclusion that the RDRPK model can predict the distance more accurately and improve the indoor positioning accuracy during the application process,which indicates that the Bluetooth positioning algorithm based on deep learning is appointed to subserve the Improvement of indoor positioning accuracy.There are 54 pictures,19 charts and 86 references in this paper.