Cross-scale defect detection method

The invention relates to a cross-scale defect detection method which comprises the following steps: S1, acquiring surface defect data of an object to be detected, and classifying and defining defects; s2, performing feature extraction on defect-containing data in the original image to obtain cross-scale defect edge features; s3, inputting the original image data and the cross-scale defect edge features of the object to be detected into the SwinIDE-merge network model, and extracting high-dimensional defect information; s4, constructing a defect detection model, outputting the high-dimensional defect information to the defect detection model, and detecting a bounding box prediction result and a classification result of the defects; and S5, aiming at the defect detection model of the cross-scale defect, adopting a Wasserstein distance as a loss function, carrying out training and weight updating on the model, and obtaining a final defect detection model. According to the feature extraction method, fewer down-sa