篇名 | Glass Defect Recognition Method Based on Improved Convolutional Neural Networks |
---|---|
卷期 | 30:6 |
作者 | Dan-Dan Zhang 、 Yong Jin 、 Bin-Yu Hu |
頁次 | 168-180 |
關鍵字 | convolution neural network 、 glass defect recognition 、 image processing 、 k-means clustering 、 parse auto-encoder 、 EI 、 MEDLINE 、 Scopus |
出刊日期 | 201912 |
DOI | 10.3966/199115992019123006013 |
The features of the same type of glass defects are quite different, thus it is difficult to accurately recognize the type of glass defects. This paper proposes an improved Convolutional Neural Networks model to solve the problem of glass defects recognition. Image processing methods are used to reduce the noise of the image, so that the edge of the defect can be clearly recognized. A sparse auto-encoder is used to learn the feature of the input samples, and trained weights are used as the convolution kernel of the network to increase the speed of recognition. K-means clustering is used instead of softmax classifier of Convolutional Neural Networks. The defect images are first clustered and then classified to improve the recognition accuracy. Experimental results show that the recognition accuracy of this method reaches up to 96%.