Research On Indoor Visible Light Localization Algorithm Based On ELM Neural Network

In recent years,with the continuous progress of information technology,people’s demand for indoor positioning services has become increasingly strong.Although the Global Positioning System has been widely used in various outdoor positioning systems,indoors,the signals from satellites are affected by obstacles such as walls,which greatly reduces the accuracy of GPS positioning and cannot meet the needs of indoor positioning.Since visible light is basically harmless to human body,it will not affect other electronic products,and at the same time,it will not be affected by electronic devices.Therefore,the use of visible light as a signal carrier for signal transmission and its application in indoor positioning has become a research hotspot for domestic and foreign scholars.The focus of this thesis is to apply machine learning to visible light indoor positioning to achieve high accuracy indoor positioning.The main work is as follows:(1)The indoor visible light localization model and the illumination distribution of LED light sources are studied.The channel characteristics under different line-of-sight links are analyzed,and the optical power distribution of the LED light source is simulated,and the noise model is also analyzed in detail.Several common indoor visible light localization methods are analyzed,and the advantages and disadvantages of several methods are compared.Several commonly used indoor localization algorithms are analyzed,and finally neural networks in machine learning are chosen as the focus of the study.(2)In this thesis,a multi-LED indoor visible light localization algorithm based on Extreme Learning Machine neural network is proposed.A multi-lamp system model with four LEDs as emitters and one Photoelectric Detector as a receiver is constructed.The simulation results show that the average localization error is 1.17 cm and the average localization time is 0.03594s.Compared with BP neural network,SVM algorithm and GA-BP algorithm,ELM algorithm not only improves the localization accuracy,but also enhances the time efficiency of the system.(3)Since the weights and thresholds of the ELM neural network are determined randomly from within the neural network,this will cause the localization system to be unstable,which will lead to the localization system easily falling into local optimum and cause the system localization error to increase.To address this problem,another localization algorithm based on Genetic Algorithm optimized ELM neural network is proposed in this thesis.The principle of GA-optimized ELM neural network is introduced in detail,and the optimized neural network is simulated and analyzed.Finally,an experimental scenario of multi-LED indoor visible light localization system is built to compare and analyze the localization performance of ELM neural network algorithm and GA-ELM algorithm in the actual environment.And the localization performance of GA-ELM algorithm and support vector machine neural network algorithm,BP neural network algorithm and GA-BP algorithm in the actual environment is compared and analyzed in detail.