Abstract: The determination of iron ore composition is very important to measure iron ore quality. The traditional method has the defects of heavy workload， low efficiency and long period. It is difficult to determine the content of all components in iron ore only by using single spectra. In the study， Anshan type iron ore was used as experimental samples. The reflectance spectra and emission spectra of the experimental samples were combined. The composition contents of samples were determined by application of spectral analysis technology. The results showed that the contents of Fe3O4， Fe2O3 and SiO2 in the samples， as well as the grade and magnetic susceptibility can be inversed by combining reflectance spectra and emission spectra. Then the model was applied to the field. The distribution of iron ore was basically consistent with the iron ore enrichment in the mining area. The average predictive residual of the grade in iron ore was 3.5%. The prediction results of the model were ideal. The method has the advantages of low working strength， conveniently， quickly， efficiently and no pollution. It can reduce the test error on the basis of further model optimization in the future. It provides a new technical method for intelligent perception of rock and mineral resources ， unmanned mining， intelligent mining and intelligent ore blending in the future.

Key words: iron ore, hyperspectral, reflectance spectra, emission spectra, inversion