Satellite cloud picture image restoration method introducing physical imaging priori knowledge constraint

The invention discloses a satellite cloud picture image restoration method introducing a physical imaging priori knowledge constraint. The satellite cloud picture image restoration method comprises the steps of S1, building a PKKernelGAN + ZSSR image degradation blurring kernel estimation deep neural network; s2, constructing an imaging degradation matrix; s3, the imaging degradation matrix is written into a loss function, a PKKernel GAN network is trained, and a cloud image degradation blurring kernel estimation network PKKernel-G is obtained; s4, constructing a'high quality-low quality 'paired cloud picture image data set; s5, training a cloud image restoration deep neural network ZSSR based on the high quality-low quality pairwise cloud image data set; and S6, inputting the original cloud image into the cloud image restoration deep neural network ZSSR trained in the step S5 to obtain a finally output high-quality restored cloud image. The physical imaging loss of the cloud image can be measured, and the im