Visual Transform-based gear pitting corrosion measurement method

The invention relates to a gear pitting measurement method based on a visual Transform. The method comprises the following steps: S1, improving a gear contact fatigue test bench, and obtaining a gear pitting image; s2, unifying the image resolution and preprocessing the image resolution; s3, making an image segmentation label and dividing a data set; s4, combining Transform with a CNN (Convolutional Neural Network), and constructing a UsViT network by using methods of residual connection, expansion convolution and the like; s5, inputting the training set into the UsViT network, and performing iterative updating training on the UsViT network by using a multi-classification cross entropy loss function; s6, constructing a relative error evaluation index of the pitting corrosion area rate; and S7, inputting the test set into the trained UsViT network, segmenting the effective tooth surface and the pitting corrosion area of the gear to obtain a pitting corrosion area rate, and evaluating the health state of the ge