Research On Multi-floor Accurate Indoor Location Method

In recent years,with the continuous development of our society,many large buildings have been built.It has become a hot research topic for many researchers how to obtain accurate location information in these buildings.At present,the existing GPS positioning is only suitable for outdoor positioning,in indoor due to signal shielding often can not achieve the desired effect.In this paper,the existing Wi-Fi network is used to extract location fingerprints in the Wi-Fi network to obtain position-related information,and the location information is put into the deep learning framework to train the neural network model with high positioning efficiency and high accuracy,so as to achieve accurate location information in large indoor buildings.Combined with the current deep neural network model,this paper designs a multi-floor indoor positioning system,which is summarized as follows:1.This paper proposes a multi-floor indoor positioning method based on extensible DNN.The method firstly improves the accuracy of data by PCA dimensionality reduction and denoising and dividing data sets,and then puts the pre-processed RSSI data into the multi-layer architecture based on extensible DNN.The architecture creatively differentiates the building,floor and exact location information by binary thermal coding,and finally obtains the information of the building,floor and exact location.2.This paper proposes a multipath and NLOS suppression algorithm based on improved CNN for multipath and NLOS phenomena in indoor positioning process.The algorithm weakens the influence of multipath and NLOS phenomena on positioning results through multi-layer convolution operation.Finally,a multi-floor indoor positioning system solution is obtained by combining the algorithm with the multi-floor indoor positioning based on extensible DNN.3.This paper designs a multi-floor positioning system,which identifies buildings and floors through the building floor identification module,obtains matching fingerprint database through the neural network training module,and obtains accurate location information through the fingerprint matching module.Finally,a practical application model is designed to transform theory into application.