Detection method based on deep learning for yellow industrial diamond

To solve problems such as low manual testing speed, high labor intensity and limited quality consistency of yellow industrial diamond test, we proposed a testing method based on deep learning for yellow industrial diamond. Firstly, a hardware system was developed to collect the data of the yellow industrial diamond by its structural characteristics. Then, the data were pre-processed by image processing, using three basic classifiers, namely VGG-16, inception-V3 and ResNet-50 to construct three network structures respectively. The information fusion of each basic classifiers and classification decision were realized by integrating fusion method. It is confirmed by verifying test that the comprehensive recognition evaluation indicators of yellow industrial diamonds with grades of 2240, 2280 and 2290 are all over 85%. This testing method is valuable and highly practicableness.