High-precision and robust visible-light dynamic imaging positioning system based on improved kernelized correlation filters and fast-weighted Levenberg–Marquardt algorithm

Abstract

This paper proposes a high-precision and robust visible-light dynamic imaging positioning system based on improved kernelized correlation filters (IKCF) and fast-weighted Levenberg–Marquardt (FWLM) algorithm. First, using IKCF, we propose a novel LED-ID dynamic recognition algorithm, which can achieve fast LED-region of interest detection, and an adaptive model update mechanism is used to reduce the interference of occlusion. Second, an FWLM positioning algorithm is designed. The direct linear transformation is used to obtain the initial estimate of the pose quickly, and LM is employed to optimize the pose iteratively by minimizing the reprojection error. An adaptive weighting matrix is used to reduce the impact of LEDs with low detection reliability on positioning accuracy. Finally, an experimental platform is set up to verify the effectiveness of the proposed dynamic positioning algorithm. The experimental results show that the average positioning time is 4.08 ms, the average positioning accuracy is 1.9830 cm without occlusion, and the average positioning accuracy is 4.4024 cm with occlusion.