High-precision automatic optical axis alignment of LED backhaul in lighting surroundings based on object detection

Abstract

Research on high-throughput optical wireless communication (OWC) has been progressing toward sixth generation. As half-power beam width of near-infrared 600 Mbps light emitting diode backhaul (LEDBH) consisting of LED and photodiode is narrow, high-precision optical axis alignment of LEDBH is important. So far, the optical axis was manually aligned using a scope. In this study, object detection using deep learning, YOLOv7, is adapted to automatically align the optical axis. Average precisions of LEDBH detection are evaluated in lighting surroundings while changing the camera exposure time. The optical axis is finely adjusted with object detection and mechanical angle adjustment.