An Iron Ore Identification Method Based on Improved Bilinear Network

In this study, the algorithm of iron ore classification based on bilinear network is discussed, which can enhance the subsequent ore processing process. Firstly, the data set has been established by collecting the graphics of three types of iron ore and utilizing the data enhancement methods. Then, a modified bilinear network model based on cross-layer feature fusion and attention mechanisms is constructed, and the ResNet-50 is introduced instead of the original VGGNet. Then, the experimental results demonstrate that iron ore graphics can be classified with a top-1 accuracy of 96.87%, confirming the network model's capacity to extract features. This research could improve the efficiency of the iron ore sorting, which the limitations of the traditional methods.