Ghost image of rotating object with group frame neural network combining frame correction algorithm

Efficient acquisition of clear images of moving objects is one of the research bottlenecks in the practical applications of ghost imaging (GI) technology. Here the representative ghost imaging of rotating objects was focused on, and a group frame neural network combined with a frame correction algorithm was proposed to promote the improvement of ghost imaging quality and efficiency for moving objects. By performing angle correction on the illumination patterns to obtain corrected bucket measurement images, the motion blur can be suppressed. These bucket measurement images are then input into the network to achieve high-quality image reconstruction. The structural similarity index measure (SSIM) of the network's output images is on average 25 times higher than traditional GI. Moreover, all the images of different angle orientations during the rotation of 360 degrees with a 1-degree step distance can be reconstructed simultaneously by making use of one frame of correlation measurement data. Therefore, the acquisition cost with the proposed algorithm can be reduced by more than 300 times compared with traditional GI without angle correction. The proposed scenario may find broad application prospects in the fields of military and remote sensing.