Chip Appearance Defect Recognition Based on Convolutional Neural Network

Sensors

Jingjing Wu¹  Xiaomeng Zhou¹  Jun Wang² 

[1] Jiangsu Key Laboratory of Advanced Food Manufacturing Equipment Technology, Jiangnan University, Wuxi 214122, China;School of Mechanical Technology, Wuxi Institute of Technology, Wuxi 214121, China;

关键词: chip appearance defects;    data cleaning;    convolutional neural network;    pattern recognition;   

DOI  :  10.3390/s21217076

来源: DOAJ

【 摘 要 】

To improve the recognition rate of chip appearance defects, an algorithm based on a convolution neural network is proposed to identify chip appearance defects of various shapes and features. Furthermore, to address the problems of long training time and low accuracy caused by redundant input samples, an automatic data sample cleaning algorithm based on prior knowledge is proposed to reduce training and classification time, as well as improve the recognition rate. First, defect positions are determined by performing image processing and region-of-interest extraction. Subsequently, interference samples between chip defects are analyzed for data cleaning. Finally, a chip appearance defect classification model based on a convolutional neural network is constructed. The experimental results show that the recognition miss detection rate of this algorithm is zero, and the accuracy rate exceeds 99.5%, thereby fulfilling industry requirements.

【 授权许可】

Unknown