A comprehensive review of defect detection in 3C glass components

Abstract With the development of consumer electronics industry, it is inevitable for the industry to use machine vision instead of manual inspection. In this paper, the defect detection of 3C glass components is summarized according to the actual production process. The defects of glass components are classified in details for the first time. The causes of these defects and the optical characteristics exhibited in the detection process are also analyzed. Because the detection effect is determined by classifier, the performance of various classifiers is discussed in details under the same criterion. On the whole, the neural network classifier is obviously better than the traditional methods, and the unsupervised classifier is better than the supervised one. The current detection accuracy is about 90%, and the computational complexity of high-accuracy classifiers is large. In the future, the improvement of accuracy and the reduction of computational complexity are still the research focus.