Neural Network Visible Light Indoor Location Based on Lambert Model Optimization

In the visible light indoor positioning system based on neural network, it is difficult to accurately locate the complex and changeable indoor environment due to the restrictive factors, such as noise interference and difference of radiation models. In this paper, the RBF (Radial Basis Function) neural network positioning system is used as a model to fit the parameter values of the real indoor wireless channel. In order to optimize the fingerprint library of the Lambert model, the multi-array visible light radiation mechanism and the RSS positioning theory are combined to reduce the difference between the actual radiation model and the theoretical radiation model. Finally, the optimized fingerprint library is used in the RBF neural network to improve the positioning accuracy. Finally, the effectiveness of the proposed algorithm is verified by the measured positioning experiment. The results show that the average positioning error after optimization is 4.68cm in the measured space of O.8m × O.8m × O.8m, and the average positioning accuracy is increased by 15.52 % compared with 5.54cm before optimization, which provides an effective scheme for improving indoor positioning accuracy.