计算机科学
光学
模式(计算机接口)
能量(信号处理)
布线(电子设计自动化)
物理
光通信
电子工程
波分复用
信号处理
拓扑(电路)
自由空间光通信
传输(电信)
衍射
高效能源利用
分束器
计算机网络
脉冲整形
多路复用
算法
耦合模理论
反射(计算机编程)
图像处理
折射率
作者
Ting He,Huan Chang,Fei Wang,Ran Gao,Xinyou Sun,Jingran Wang,Yue Pan,Qi Xu,xiaolong Pan,Dong Guo,zhipei li,Wei Cui,Yinglong Wang,Xiangjun Xin
出处
期刊:Optics Letters
[Optica Publishing Group]
日期:2026-07-02
卷期号:51 (14): 3805-3805
摘要
Orbital angular momentum (OAM) mode classification is challenging for single-layer phase-only diffractive neural networks due to the complex phase structure and spatially extended energy distribution of OAM beams, which are difficult to map into well-separated continuous output patterns under limited modulation degrees of freedom. To address this mismatch, we propose a binary-spot-mapped single-layer diffractive neural network (BS-D2NN) to realize region-based energy routing for OAM mode classification, which provides a compact all-optical energy-routing solution with low system complexity and direct region-based optical readout. Experimental results show that the proposed BS-D2NN achieves an overall classification accuracy of 93.89% under turbulence-free conditions, while effectively routing optical energy into class-specific target regions for mode classification and spatial demultiplexing. Supplementary experiments further indicate that the same framework remains applicable under turbulence conditions through condition-specific retraining, without changing the underlying network architecture or optimization principle.
科研通智能强力驱动
Strongly Powered by AbleSci AI