Highly adaptive single-shot surface shape distortion measurement method for high-power lasers

Real-time online monitoring of the surface shape distortion (SSD) of the deformable mirror (DM) is essential to the design, control, and performance of high-power laser facilities. Here, we report a single-shot highly adaptive SSD measurement method based on deep learning for surface shape changes measurement of the DM in high-power lasers. This method could adapt to the narrow cavity spaces in high-power laser facilities and enable fast surface shape reconstruction by utilizing a reflected fringe image and a trained network. The deep learning network, using a fully attentional transformer-based backbone for long-range dependency modeling, enhances surface shape reconstruction speed and precision by focusing on global feature extraction and reconstruction over local convolutional features. The results of the simulation and experiment show that the method could accurately predict the surface shape changes of DM in high-power lasers, indicating high speed and strong adaptability in the SSD reconstruction.