Industrial CT defect detection method based on deep learning

Abstract

The invention relates to an industrial CT defect detection method based on deep learning. The method comprises the following steps: collecting data; dividing a data set; labeling the image; performingimage enhancement; constructing a model; customizing a loss function and an evaluation standard; training a model: training the model and storing the best weight of the trained model; post-processing: inputting the verification set picture into the model to obtain an original defect mask corresponding to the verification set original image, and performing subsequent processing on the mask; calculating the area of a defect region: calculating the area of a defect pixel through the obtained mask image containing the defect; defect segmentation: loading the weight stored in the training model asa prediction model, inputting the model to obtain an original mask image of which the original image contains various defects, and performing post-processing and defect region area calculation on themask image to obtain a final defect-containing mask image and a defect area; according to the invention, the defect area can be accurately and rapidly detected and identified in actual production.

Description

technical field [0001] The invention belongs to the field of artificial intelligence, and in particular relates to an industrial CT defect detection method based on deep learning. Background technique [0002] In the current field of industrial defect detection, most still use the method of visually observing the defect to determine the defect area and manually mark it. This method cannot determine the precise defect area and consumes a lot of time; another method is to use traditional image processing. Technologies, such as SIFT+SVM, cluster the extracted image features and then classify them, and in recent years, VGG or DenseNet classification models based on deep convolutional neural networks, and instance segmentation models based on Mask RCNN, etc., but this classification method is aimed at When the foreground and background of defects are unbalanced or the types of defects are unbalanced, the accuracy of the model is not good, which leads to the low accuracy of defect...

AI Technical Summary

Problems solved by technology

[0002] In the current field of industrial defect detection, most still use the method of visually observing the defect to determine the defect area and manually mark it. This method cannot determine the precise defect area and consumes a lot of time; another method is to use traditional image processing. Technologies, such as SIFT+SVM, cluster the extracted image features and then classify them, and in recent years, VGG or DenseNet classification models based on deep convolutional neural networks, and instance segmentation models based on Mask RCNN, etc., but this classification method is aimed at When the foreground and background of defects are unbalanced or the types of defects are unbalanced, the accuracy of the model is not good, which leads to the low accuracy of defect area detection in actual production and increases the production cost

Images