A Novel Three-dimensional Indoor Localization Algorithm Based on Visual Visible Light Communication Using Single LED

Visible light positioning (VLP) is widely believed to be a cost-effective answer to the growing demand for Indoor positioning. However, because of the nonlinear and highly complicated relationship between 3D world coordinate and 2D image coordinate, there is a need to develop effective VLP location algorithm to locate the positioning terminal using image sensor. Besides, due to the high computational cost of image processing, most existing VLP systems fail to deliver satisfactory performance in terms of real-time ability and positioning accuracy, both of which are crucial for the performance of indoor positioning system. The field of view (FOV) of image sensor affects the number of LEDs. Therefore, this paper proposes an image-sensor-based single-light positioning system and sets up relevant experiments to test the proposed system. What’s more, the real-time ability is taken into consideration, which greatly improves the robustness and practicality of the system. As for the ID information of each LED, it utilizes the rolling shutter mechanism of the Complementary Metal Oxide Semiconductor (CMOS) image sensor and combines machine learning algorithm to identify. Different from the traditional LED-ID modulation and demodulation methods, the LED-ID detection and recognition problem was treated as a classification problem in machine learning filed. In this paper, two typical linear classifiers in the case of binary classification problem are introduced. As the features of different LED-ID is linear separable in the feature space, linear classifiers are used to identify the LED-ID. The scheme proposed could improve the speed of LED-ID identification and the robustness of the system by off-line training for the classifiers and online recognition of LED-ID. When there’s only one LED in the camera image, it means the distance between the terminal and LED is so close that the diameter of the LED can be measured. Therefore, it is able to use the proposed single-light positioning algorithm to achieve a high precision positioning and to solve the problem of limited number of LEDs in FOV in the existing literature. The proposed single-light positioning algorithm also has lower calculation complexity. Experimental results show that the proposed single-light positioning algorithm provides an accuracy of 7.39 cm and the computational time is 43.05 ms, which reflects a good performance in both accuracy and real-time ability.