Design and Implementation of Few-Mode Fiber Mode Decomposition and Beam Quality Measurement System Based on Deep Learning

Abstract

This paper presents the design and implementation of a deep learning technology-based characterization and analysis system for the output beam of a few-mode fiber, which is used for real-time monitoring of spot-related parameters in laser application systems. Based on an improved Convolutional Neural Network (CNN) model, a simple and user-friendly system is developed using Python language and Django framework, which realizes the functions of high-speed and automated acquisition of light spot images, real-time mode decomposition of few-mode fibers, real-time measurement of beam quality, and storage of historical analysis data. It supports rapid and automated analysis of few-mode fibers, enabling users to effortlessly perform modal decomposition and real-time measurement of the M²factor. Through experimental tests, we have verified the effectiveness and accuracy of the system in dealing with few-mode fibers containing three intrinsic modes, which provides an important application value in the field of optics.