【摘要】 分振幅光偏振仪是一种测量速度快的光波传感器,而数据处理是分振幅光偏振仪的应用基础,针对标准神经网络存在的缺陷,提出一种改进神经网络的分振幅光偏振仪数据处理方法。采用分振幅光偏振仪的电信号作为输入,入射光斯托克斯参数作为期望输出,采用神经网络拟合输入与输出之间的关系,K聚聚类算法对神经网络参数进行优化,对分振幅光偏振仪数据处理的测试实验结果表明,本文方法获得了较高精度的入射光斯托克斯参数估计结果,性能要优于传统方法。更多还原

【Abstract】 The division of amplitude photopolarimeter is an optical sensor with fast measuring speed. Data processing is the application basic of division of amplitude photopolarimeter. A data processing method for division of amplitude photopolarimeter with modified neural network is proposed to solve defect of standard neural network. Electric signal is taken as input while stokes parameters of the incident light are used as the expected output,the relationship between input and output is fitted by neural network,and the parameters of neural network are optimized by K-means algorithm,data processing test results for division of amplitude photopolarimeter show that the method can obtain higher accuracy of the stokes parameters estimation results,performance is superior to traditional division of amplitude photopolarimeter data processing methods.更多还原