Chip defect recognition method based on convolutional neural network

Abstract

The invention belongs to the field of image processing, and specifically discloses a chip defect recognition method based on a convolutional neural network (CNN). The method comprises the following steps of: acquiring respective types of chip images according to preset classification types, and performing size normalization to construct a data set; constructing a CNN initial model including an input layer, a convolution layer, a downsampling layer, a full connection layer and an output layer; randomly extracting images from the constructed data set to construct a training set, and inputting the images in the training set into the constructed CNN initial model for training to obtain a CNN final model; acquiring a chip image to be recognized as the input layer to be directly input into the final CNN final model, and completing the image feature extraction by successively passing the chip image through the convolution layer, the downsampling layer and the full connection layer, and enabling the output layer to output a recognition result so as to complete online chip defect recognition. The method can realize the rapid recognition of chip defects, has high recognition accuracy and high speed.

Description

technical field [0001] The invention belongs to the field of image processing, and more specifically relates to a chip defect recognition method based on a convolutional neural network. Background technique [0002] Chip defect detection is crucial to controlling the quality of chip products. The currently used chip defect recognition method adopts the method of "image preprocessing-target segmentation-feature extraction-defect recognition and classification". This recognition method has no learning ability, takes a long time to identify, and relies heavily on models. The feature extraction and description are complicated and cumbersome, requiring a lot of professional knowledge of the operator, requiring high experience of the operator, and not being adaptable to the environment. Therefore, the traditional chip defect recognition method can no longer meet the requirements of defect recognition rate and recognition speed in the modern chip manufacturing field, and research ...

AI Technical Summary

Problems solved by technology

[0003] Aiming at the above defects or improvement needs of the prior art, the present invention proposes a chip defect recognition method based on a convolutional neural network. This method does not need to perform complex preprocessing on the image, and directly inputs the gray value information of the chip image into the volume In the convolutional neural network model, the convolutional neural network is used to extract image features layer by layer to realize the identification of chip defects. The convolutional neural network model constructed by the present invention has strong robustness. Through the study of the data set, it can continuously Improve the accuracy and speed of recognition, with the advantages of high recognition accuracy and fast speed, and can be used to identify unqualified chips with apparent defects in the chip manufacturing process

Images