Indoor visible light positioning based on fusion of subregion BES-ELM and WDME weighted dual-mode

Aiming at the problems of low indoor positioning accuracy and large boundary area positioning error, an indoor visible light positioning method based on the bald eagle search-extreme learning machine (BES-ELM) neural network and weighted dual-mode edge (WDME) positioning model was proposed. In this method, a visible light system structure with a single LED and five photodetectors was proposed, and the room was divided by fuzzy c-means clustering algorithm. The BES was used to optimize the ELM neural network, and the BES-ELM positioning model was established in different regions. Aiming at the boundary area, a weighted dual-mode edge (WDME) positioning model was constructed to achieve accurate edge location. Based on the indoor environment simulation of 3.2 m×3.2 m×3 m, the results show that using the BES-ELM algorithm to locate the center area, the average positioning error is 0.011 7 m, and the minimum positioning error is 0.001 9 m. Using the WDME positioning model to locate the edge area, the average positioning error is 0.013 3 m, which is 84%, 27%, and 26% higher than that of ELM, Elman and BES-ELM models, respectively. Therefore, the proposed visible light positioning method reduces the overall area positioning errors, especially improving the positioning accuracy of edge area.