改进型RBF神经网络室内可见光定位方法及系统

The invention discloses an improved RBF neural network indoor visible light positioning method and system, in which the light signal emitted by the LED in the indoor space passes through the indoor space channel, and the received end uses a PD detector to convert the obtained light signal into an electrical signal for positioning, including calculating the light intensity value with the least fluctuation of the optical power, calculating the positioning and optimizing the RBF neural network. The optimization of RBF neural network includes: using KPCA-K-Means ++ model to improve the performance of RBF neural network, that is, using KPCA-K-Means ++ model to cluster the RSSI data collected at the receiving end, and obtain the optimal cluster center and the number of clusters as the number of neurons and neuron centers in the hidden layer; The GA-LMS model is established to optimize the parameters of RBF neural network, that is, accurate width and connection weight are obtained by using GA-LMS, and accurate positioning coordinates are finally obtained, and the invention has stronger generalization.