Indoor Visible Light Positioning for a Single Partially Visible LED

With the IEEE 802.11bb standard being adopted to incorporate light communication, research interest in visible light positioning (VLP) has increased. While several VLP techniques have seen success using complementary metal oxide semiconductor sensors for a single LED, in most cases, the field of view (FoV) of a front camera on a smartphone is much smaller than the rear camera, and lights are placed sparsely in offices since their primary objective is illumination. Hence, during indoor navigation, the front camera is bound to encounter far more partial images of the light than complete images. The proposed technique seeks to solve this problem by performing positioning on images where only two corners of a square light are in the FoV. While most offices have square or rectangular panel lights, we have chosen to use square lights owing to the added difficulty in positioning arising from all sides being equal in length. We detect the corners of the light from an image and order them based on inertial measurement unit (IMU) readings from smartphones to perform structure-based positioning. The proposed technique achieved 2.27 cm average 3-D positioning error on a partial light image dataset captured at two different heights.