Chip surface defect parallel detection method based on improved convolution variational auto-encoder

Abstract

The invention discloses a chip surface defect parallel detection method based on an improved convolution variational auto-encoder. The problem that chip surface defect detection still needs to be intelligent in the prior art is solved. The method comprises the following steps: step 1, acquiring and preprocessing a chip image; step 2, constructing and training an improved convolution variational auto-encoder; and step 3, performing real-time defect detection on the chip to be detected. According to the method, a traditional variational auto-encoder is partially improved, all defects different from those of a normal sample can be detected only by using the normal sample for unsupervised training and without labeled defect sample data, manpower and material resources are reduced, meanwhile, the original image is decomposed into a plurality of sub-images by the method and then the sub-images are processed in parallel, and the detection accuracy is improved. And various defects on the surface of the chip can be efficiently and quickly detected at low cost.

Description

technical field [0001] The invention relates to the technical field of computer vision applied to chip screening, in particular to a method for parallel detection of chip surface defects based on an improved convolutional variational autoencoder. Background technique [0002] The purpose of chip surface defect detection is to eliminate defective chips with surface defects such as character defects, pin defects, and surface scratches on the production line, which is the key to controlling product quality. Traditional defect detection methods based on explicit feature extraction require manual design for feature extraction, which is time-consuming; while most current surface defect detection methods based on deep learning are based on supervised learning of a large number of defect samples, requiring a large number of well-labeled training Samples are learned end-to-end, but it is very difficult to collect a large number of defect samples. The accidental and random nature of d...

AI Technical Summary

Problems solved by technology

[0003] The invention improves the problem that chip surface defect detection still needs to be intelligent in the prior art, and provides a chip surface defect parallel detection method based on improved convolutional variational autoencoder with short detection time and comprehensive detection

Images