Rapid, Non-destructive Identification of Iron Ores-Based Random Forest (RF) Using Visible and Near-Infrared Spectroscopy

Visible and near-infrared (Vis–NIR) spectroscopy technology can better meet the needs of identification of mineral differences in iron ores by using a large amount of chemical bond vibration information contained in the spectra. In this paper, the original Vis–NIR spectra data of six kinds of iron ore are dimensionally reduced by the principal component analysis (PCA) algorithm, and the influence of sample noise and outliers is eliminated at the same time. After processing, the sum of the contribution rates of the first three principal components of the reflectance spectra reaches 99.27 pct, and the eigenvalue inclusion rate of the principal component analysis data is high, which has a good data processing effect. Then, based on the random forest (RF) algorithm, the spectra data after PCA are classified and predicted. After multiple authentications, the identification accuracy of the algorithm can reach more than 95 pct on average. Finally, the receiver operating characteristic (ROC) curve is used to characterize the prediction success rate of the recognition model and its accuracy for the recognition of single iron ore.