基于神经网络的消色差旋转双棱镜反向解算

Inverse calculation of achromatic Risley prism based on neural network

  • 摘要: 为了实现基于双棱镜的光电跟踪平台,需要根据光束指向精确求解两棱镜转角。基于非近轴光线追迹法与二步法,利用神经网络拟合了消色差旋转双棱镜角度差值与偏转角之间的关系,并在求解方位角与消色差旋转双棱镜两组棱镜转角的时候,将其中的非线性关系与线性关系分开,利用神经网络对其中的非线性关系进行了拟合,最终得到了消色差旋转双棱镜转角与出射光束指向的数值关系。实验表明,在仅用3层神经网络共20个神经元的条件下,得到了两棱镜旋转角度反向解算值达到了 0.000 1°量级。

     

    Abstract: To achieve an optoelectronic tracking platform based on biprisms, it is necessary to accurately calculate the angle of the Risley prisms based on the beam direction. This article uses non-paraxial ray tracing method and two-step method to fit the relationship between the angle difference and deflection angle of achromatic rotating biprism using the Neural Network. In solving the azimuth angle and the angle of achromatic rotating biprism, we separate the nonlinear relationship from the linear relationship, which is also fitted by using neural network. Finally, the mapping relationship between the angle of achromatic rotating biprism and the direction of the outgoing beam is obtained. The experiment shows that the exact inverse calculation of angle value of 0.000 1° is obtained under the condition of using only three layers of neural network with 20 neurons.

     

/

返回文章
返回