Misalignment Calculation on Off-Axis Telescope System via Fully Connected Neural Network

Abstract

On the process of alignment for pupil off-axis telescope system, the determination of decentering and tilt misalignments is a key step. In this paper, we proposed a new method which adopts the fully connected neural network (FCNN) as a fitting tool to establish the nonlinear mapping relation between misalignments of different fields of view (FOVs) and Zernike coefficients. Firstly, we establish a pupil off-axis reflection telescope model, then decentering and tilt misalignments are introduced to acquire corresponding aberrations that represented by Zernike coefficients. We use aberrations as the inputs of FCNN; misalignments and FOV as outputs. FCNN is trained by the combination of inputs and outputs as a dataset, and we use a new dataset to test the accuracy and effectiveness of the trained FCNN. The results show that the mean absolute error (MAE) of the X-axis decentering error, V-axis tilt error and angle of view are 0.0506 mm, 0.0204° and 0.0124°, respectively. These results demonstrate that the proposed method is effective and feasible to calculate the misalignments.

Keywords

• Off-axis telescope
• misalignment
• decentering and tilt error
• Zernike coefficients
• fully connected neural network