基于KPCA-K-means++和GA-LMS模型的改进RBF神经网络室内可见光定位

Aiming at the nonuniformity of the received optical power and low positioning precision in the indoorvisible light positioning，a received signal strength indicator (RSSI) visible light positioning method is proposedbased on adaptive flower pollination quantitative light source optimization scheme combined with improved radialbasis function (RBF) based neural network. The adaptive flower pollination algorithm optimizes the light intensityof the transmitter processing the received uniform optical signal using an improved RBF-based neural network RSSIpositioning method，resulting in accurate and effective positioning. The kernel principal component analysis K-means++(KPCA-K-means++)clustering model is used to preprocess the received RSSI sample value.Theoptimal cluster number and cluster center are obtained as the number and central value of the hidden layer neurons.The genetic algorithm and least mean square (GA-LMS) model is used to optimize the parameters of the RBF neuralnetwork. According to simulation results，the received optical power ranges from -28.6 dBm to -25.1 dBm in anindoor space of 9 m×12 m×3.5 m. Moreover，the positioning error is less than 0.l m. Therefore，the proposedimproved visible light positioning method has higher positioning accuracy and stronger practicability advantages.