High precision indoor positioning method based on visible light communication using improved Camshift tracking algorithm

Abstract Recently, visible light communication (VLC) has been widely used in indoor positioning, which has such advantages as cost-saving, environmentally friendly, anti-radio frequency interference (anti-RF-interference) and so on. Considering that existing VLC-based indoor positioning systems suffer from such obstacles as blur effects and shield effects, a precise robust positioning method using improved Continuously Adaptive Meanshift (Camshift) tracking algorithm is proposed to locate the positioning terminal. The improved Camshift algorithm is used to track the region of interest (ROI) of the LED to improve the robustness of the visible light positioning (VLP) system. Classical Camshift algorithm build an one-dimensional histogram with Hue component from HSV (hue, saturation, value) color space, which may cause tracking failure when illumination variation or similar color interferes occur. Therefore, the proposed algorithm utilizes hue and saturation components from HSV space to build 2-Dimensional (2-D) color feature histogram. The H and S components of the HSV are combined with the Camshift algorithm to improve tracking accuracy and robustness. What is more, even most parts of light-emitting diode (LED) are shielded or broken, high-precision tracking can still be achieved by recognizing the color features and local detection regardless of shape change, thereby ensuring the positioning robustness. Furtherly, shield effects and background interferences are introduced to simulate actual positioning scenes. Experimental results show that the proposed algorithm can provide an average accuracy of 0.95 cm and ensure that 90% of total tracking error is less than 1.79 cm, indicating that the LED tracking is so accurate that the positioning accuracy of the positioning algorithm is not affected by the tracking algorithm. Meanwhile, the average computing time of the tracking algorithm is 0.036 s for per frame, which demonstrates that the computational cost required for Camshift is so small that the positioning algorithm is not affected by the tracking algorithm and can still have good real-time performance. Therefore, the proposed algorithm has broad application prospects in fields of dynamic positioning and tracking services.