Fusion-driven visible light positioning using photodiode and event camera.

Abstract

Visible light positioning (VLP) serves as a critical technology enabling high-accuracy indoor sensing applications. However, photodiode (PD)-based and CMOS-based VLP systems are respectively susceptible to ambient interference and limited dynamic range of sensitivity. To address these challenges, what we believe to be a novel hybrid VLP framework that integrates an event camera and a PD is proposed. Specifically, the extended Kalman filter (EKF) fuses PD and camera data to leverage their respective strengths to achieve three-dimensional (3D) positioning with a single LED. Moreover, a dynamic compensation algorithm is proposed to improve the second-order radial distortion of the camera and enhance the positioning accuracy. Extensive experiments demonstrate that the proposed VLP system achieves an average localization accuracy of 1.98cm in horizontal 2D plane and 0.83cm in the vertical direction. Furthermore, the positioning robustness of the proposed VLP system in low signal-to-noise ratio scenarios is verified by blocking the LED and introducing ambient light interference.