In the process of digital detection operation of the vessel scale in the port, the digital observation of water scales may have the problem of incomplete numbers due to the influence of adverse factors such as reflection and hull corrosion. Based on this, this paper proposes a method for repairing and identifying the number of the vessel ruler using deep learning, including proposing a new loss function and the improvement of the network model by context mixing and module, so as to achieve the repair effect and recognition accuracy of the water scale image that is satisfactory. The experimental results show that the image restoration method of GAN based on the water ruler can achieve efficiently the restoration of edge contours and feature textures, and the accuracy rate is greatly improved compared with the confidence, which provides great convenience for the port water scale viewing and detection.