Large dynamic range of secondary position errors detection for the coaxial three-mirror system using dual-branch convolutional neural network

When the space optical remote sensing system works in orbit, it is easy to be affected by the external environment such as vibration and shock, space environment and optical system itself. In this case, the position of the secondary mirror relative to the primary mirror will shift, which causes wavefront aberrations and deteriorates the image quality. The traditional position errors detection technique has the following disadvantages such as complex device, time-consuming calculation, low accuracy and small dynamic range. In view of the shortcomings of the existing secondary mirror position errors detection technology, a large dynamic range secondary mirror position errors detection method based on two-branch convolutional neural network (CNN) is proposed. Taking the coaxial three-mirror system as the research object, its maximum RMS value of offset wavefront aberration is above 6 λ. This method uses a new dual-branch CNN to establish the nonlinear relationship between symmetrical defocus point spread function (PSF) and the secondary mirror position offset under a single field of view, which improves the dynamic range of secondary mirror position errors detection. The prediction effects of the network trained with different fields of view data are compared, and the effects of field deviation and environmental noise on the prediction accuracy of the network are analyzed. The simulation results show that the proposed method can effectively improve the dynamic range of secondary mirror position errors detection, also has high sensing accuracy, and has good generalization ability.