Deep Learning Based Detection of Anti-Reflective Obstacles in VLC Systems

In Visible Light Communication (VLC) systems, the existence of non-reflective obstacles in the field of view of the receiver may cause deterioration or complete blockage of the optical signal which leads to an error-prone decoding process. We observe that image processing techniques, in combination with deep learning, provide effective ways to deal with this problem. In this work, we propose an efficient technique for the decoding of the signal in frames with partially occluded areas. First, a deep learning detector is used for the extraction of the obstacle, resulting in the determination of the regions of interest in each frame. Then, in a second step the decoding of the signal is employed, based on image processing techniques. The proposed method is evaluated in terms of the size of the obstacle and it is compared with other decoding schemes, providing very promising results.