For the development and utilization of Rosaceae plant resources, it is of great significance to collect information on different Rosaceae plant species and clarify their family and generic relationships. In this study, leaves, petals, and stamens of different Rosaceae plant species are analyzed through Fourier transform infrared (FTIR) spectroscopy combined with principal component analysis (PCA), hierarchical cluster analysis (HCA), and soft independent modelling of class analogy (SIMCA). The results revealed that the leaves, petals, and stamens of Rosaceae contained polysaccharides, proteins, lipids, calcium oxalate, lignin, and other components, while the petals and stamens contained phenols in addition. The FTIR spectra of different types of leaves are found to be similar, but the absorption peak intensities in the range of 1660?1000 cm^-1 differed significantly. Upon using this range for PCA, the first two principal components could achieve more than 97% of the cumulative variance contribution rate. Using HCA, 11 species of the plant could be correctly classified at the subfamily level. Combined with the SIMCA discriminant model, in the classification of Rosaceae plants with different leaves, petals, and stamens, the correct classification rate reaches 96.08% with the full spectral data in the range of 4000?400 cm^-1, and 100% accuracy can be achieved with the data in the range of 1800?800 cm^-1. The results reveal that FTIR spectroscopy combined with statistical analysis and discriminant modeling is a suitable method for accurately classifying different species of Rosaceae plants at subfamily and genus levels.