Inspection Method for Appearance of Underwater Structure of Wharf

This paper pre-trains the adaptive super-resolution model based on meta-transfer learning through external image data sets, so that the model can learn the prior information of image reconstruction; random Gaussian sampling is used to randomize the images in the external image data set The down-sampling of parameters makes the training data in the meta-migration learning process contain multi-task information; down-sampling the target low-resolution image to be reconstructed to obtain low-resolution sub-images as training data to train the model; relying on the twin neural network and The model is trained in confrontation, and the model parameters are adjusted to complete the training of the model by comparing the difference between the low-resolution image and the low-resolution sub-image; the model is used to reconstruct the super-resolution image on the target image. Finally, the feasibility of the method is verified through experiments.