Abstract
Accurate gemstone classification is critical to the gemstone and jewelry industry, and the good performance of convolutional neural networks in image processing has received wide attention in recent years. In order to better extract image content information and improve image classification accuracy, a CNNs gemstone image classification algorithm based on deep multi-feature fusion is proposed. The algorithm effectively deeply integrates a variety of features of the image, namely the main color features extracted by the k-means++ clustering algorithm and the spatial position features extracted by the denoising convolutional neural network. Experimental results show that the proposed method provides competitive results in gemstone image classification, and the classification accuracy is nearly 9% higher than that of CNN. By deeply integrating multiple features of the image, the algorithm can provide more comprehensive and significant useful information for subsequent image processing.