Research On Indoor Visible Light Channel Source Layout And Location Algorithm

With the continuous development of indoor positioning technology,the accuracy of traditional GPS can no longer meet the requirements of high precision positioning in indoor environment.Therefore,some technologies such as Bluetooth positioning are derived.Although these technologies can achieve basic positioning,these technologies are subject to high electromagnetic interference,low positioning accuracy and occupy electromagnetic spectrum resources.Visible light positioning based on visible light communication has great development potential.However,at the present stage,there are still some problems that need to be solved for indoor visible light positioning technology: indoor visible light channel modeling in complex indoor environment;Optimize the position,Angle,power and other information of the lamp source under the indoor channel model;Improve the flatness of receiving surface illumination and signal-to-noise ratio;Improve positioning accuracy by positioning algorithm.In this paper,a parallel fully connected convolutional neural network combined with Monte Carlo method is proposed to establish a prediction model of illumination flatness parameters from source coordinates,power and orientation angles to the receiving surface.Monte Carlo method is used to calculate the flatness of the intensity distribution of the receiving surface.The optimized particle swarm optimization algorithm was used to obtain the optical power distribution of the optimal receiving surface,the light source layout and the light source power distribution under the signal and noise score layout.Computer vision combined with visible light positioning is used to improve positioning accuracy.The specific work is as follows:(1)In an indoor environment of 5m×5m×3m,the light intensity distribution on the receiving surface was sampled by adjusting the information of the lamp source,and the fingerprint database of the light intensity distribution model was built.Monte Carlo method was used to calculate the flatness of the light intensity distribution.The generative adversarial network is used to augment the fingerprint database data,and the simulated fingerprint database data set is generated to expand the sample.(2)A parallel fully connected neural network channel model was proposed to predict the light intensity distribution of the receiving surface from the light source information,and a parallel fully connected convolutional neural network flatness calculation model was proposed to predict the light intensity distribution flatness of the receiving surface from the light source information.The light intensity distribution model fingerprint database was used to train the model.(3)The parallel fully connected convolutional neural network channel model is taken as the objective function to be optimized,and the initial population of driven particle swarm optimization algorithm is optimized using K-Means++ to obtain the light source layout information under the optimal flatness parameter.(4)The computer vision algorithm is used to propose the optical path priority redundancy positioning algorithm and visual priority redundancy positioning algorithm respectively.The target detection algorithm is optimized and combined with the perspective algorithm to obtain the position information for visible light positioning redundancy positioning,realizing the multi-mode indoor positioning of optical communication and pattern recognition,and ensuring the robustness of target positioning.