| abstract |
The invention discloses a surface defect detection system, comprising an image acquisition module and a U-Net model; the image acquisition module is used to acquire an image of a workpiece, preprocess the image, extract the original image data of the workpiece, and convert the original image of the workpiece The data is filtered, and the filtered image data is input into the U-Net model. The U-Net model includes an encoder. The U-Net model compresses the image and captures key features, and the encoded image is sampled back by the encoder. Original size, and obtain key information to optimize the custom loss function, and output a binary mask with the same size as the original image; perform binary classification through convolutional network, and output U‑Net model to process data; According to U‑Net model The processed data is compared with the original image data of the workpiece to judge the defect information of the workpiece. |