Chip surface defect classification device and method based on generative adversarial network

Abstract

The invention discloses a chip surface defect classification method based on a generative adversarial network. The method comprises the following steps: S01, inputting a random vector into an image processing module, obtaining a composite image, combining a photographed image and the composite image into an image set, and enabling the image set to comprise a training image set; s02, inputting thetraining image set into a training unit for model training to obtain a recognition model; and S03, inputting a to-be-recognized image into the recognition model, and obtaining a corresponding defect type through recognition of the recognition model. According to the chip surface defect classification device and method based on the generative adversarial network provided by the invention, the synthetic image is generated by using the random vector, and the synthetic image is used as one part of the image set, so that a large number of training images and test images are formed, and the trainingaccuracy and generalization ability of a recognition model are improved.

Description

technical field [0001] The invention relates to the field of machine vision, in particular to a chip surface defect classification device and method based on a generative confrontation network. Background technique [0002] Integrated circuit chips are widely used in various fields and are the key to national economic development and information security. However, during the manufacturing process of its package, the defects generated on the chip surface will directly affect the working life and reliability. Traditional manual detection and classification methods have shortcomings such as relying on subjective experience, time-consuming and labor-consuming, and high false detection rate, which can no longer meet the needs of high-precision and high-speed production lines. [0003] The traditional chip defect classification technology based on machine vision has a relatively mature framework system, which includes the following processes in turn: image acquisition, defect ima...

AI Technical Summary

Problems solved by technology

However, in actual production, due to the limited number of defect images generated online, it is difficult to meet the large demand for deep convolutional neural network training data; if the number of defect images used for training is insufficient, the accuracy of the model will be affected

Traditional image enhancement methods based on captured image processing and changes only generate new images based on limited invariance, and the generalization ability is weak, which cannot effectively improve the classification performance of the recognition model

Images