A Localization Method Using Reflected Luminance Distribution

Abstract

Visible light positioning is expected to become an effective means of indoor localization, but most existing methods require the capture of direct light, which is a significant limitation. In this paper, we propose a novel localization method based on received signal strength, which does not need to use direct light signals but instead uses reflected light from the floor. Our method is based on two observations. First, assuming a flat floor, the reflected light from the light source decays according to a gradient model whose peak is just below the light source. Second, the decay can be estimated effectively even for a floor surface several meters away from the light source. Inspired by these observations, we propose a method for estimating the coordinates of the floor surface and the two-dimensional coordinates of the light receiver (a camera). The method uses pattern matching within the distribution of signal decay measurements obtained via photographs of an arbitrary floor surface. Although the proposed method is vulnerable to shadow effects such as those caused by the camera tripod used in our experiments, we achieved a 90th percentile of less than 32 cm in our offline experiments. After removing the tripod shadows from the captured video manually, the same technique achieved a 90th percentile of 22 cm. To investigate the efficiency of the pattern matching, we also conducted experiments on the relationship between pixel utilization and localization results. In this paper, we also discuss camera posture estimation and power consumption issues.