Synthetic picture generation method and device, electronic equipment and storage medium

Abstract

The invention relates to the technical field of industrial defect detection, in particular to a synthetic picture generation method and device, electronic equipment and a storage medium. The syntheticpicture generation method comprises the following steps of: segmenting an input picture to obtain segmented pictures, wherein the adjacent segmented pictures have a partial overlapping region; numbering the segmented pictures, wherein serial numbers and the segmented pictures form numbered picture information pairs; adding random noise as input of a generator, and training a generative adversarial network by utilizing the segmented pictures to finally generate a plurality of defective pictures; predicting the serial numbers corresponding to the defective pictures according to the numbered picture information pairs; and performing picture splicing according to the defective pictures corresponding to a predicted number index to obtain a synthesized picture. According to the synthetic picture generation method provided by the invention, the vivid synthetic picture can be generated, the picture style is effectively expanded, and the problems of limited expansion style and macroscopic boundary of the generated picture in the existing expansion method are solved.

Description

technical field [0001] The invention relates to the technical field of industrial defect detection, in particular to a method, device, electronic equipment and storage medium for generating a composite image. Background technique [0002] In industrial production, factories usually set up cameras to monitor the quality of industrial products. However, in industrial defect detection, there are very few defect sample pictures, and the defects are small, and the pictures taken are large. It is difficult for deep learning methods to augment this dataset. Traditional data augmentation methods use rotation, cropping, and combination methods. The rotation and combination methods are limited, and the images obtained by cropping are just a drop in the bucket. At present, the generation confrontation network can generate a large number of realistic pictures, and the boundary of the original picture can be extended by the generation confrontation network to obtain high-resolution pict...

AI Technical Summary

Problems solved by technology

[0004] The traditional data augmentation method has limited image styles, and the effect of augmenting high-resolution images is not good; the generative confrontation network generates high-resolution images, and the boundary between the center of the image and the extended boundary is visible to the naked eye

Images