Zero-sample ISAR image super-resolution method, device and equipment based on meta-learning

The invention relates to a zero-sample ISAR image super-resolution method, device and equipment based on meta-learning. The method comprises the following steps: initializing an ISAR image super-resolution network and a fuzzy kernel prediction network according to the size of a high-resolution ISAR image, a super-resolution scaling multiple and the size of a to-be-predicted degradation fuzzy kernel; the ISAR image super-resolution network and the pasty kernel prediction network respectively carry out iterative computation on the initially input random noise, an ISAR reconstructed image and a degeneration blurring kernel are respectively generated after each iteration until the generated ISAR reconstructed image and the degeneration blurring kernel converge, and a high-resolution ISAR image is obtained. And after each iterative calculation of the ISAR image super-resolution network and the paste kernel prediction network, the ISAR image super-resolution network and the paste kernel prediction network are subje