A Deep Learning-Assisted Visible Light Positioning Scheme for Vehicles With Image Sensor

Abstract

Visible light communication (VLC) based on image sensor (IS) as the receiver is one of the supplementary technologies of radio frequency communication. By combining with image processing technology, VLC has various applications in smart home, underwater communication and intelligent transportation system (ITS). In the paper, it focuses on the vehicle positioning service provided by VLC in ITS. For the weak image spatial separability caused by long-distance VLC transmission between the vehicle and the infrastructure, it will lead to serious deterioration in communication and positioning performance. Thus, a deep learning-assisted IS-based visible light positioning (VLP) scheme is proposed and experimentally demonstrated. At the transmitter, a novel coding method is proposed to support both short-distance and long-distance communication based on VLC in ITS. Meanwhile, it can meet the dimming requirements of LED road infrastructure. In addition, in the proposed VLP scheme, it uses an artificial neural network (ANN) to predict the location of the vehicle. It is demonstrated that long-distance communication, high-accuracy positioning, and LED dimming can be realized simultaneously. The results show that, by using the proposed scheme, as the transmission distance is 2 m, the bit error rate (BER) of system is 1.25 × 10−4. And the average positioning error is 19.8 mm at the maximum simulated distance of 30 m.

Keywords

• Visible light communication
• vehicle positioning
• image sensor
• artificial neural network