The surface particle size and distribution characteristics of celestial bodies (e.g., the Moon, asteroids, etc.) will affect the interpretation of hyperspectral remote sensing data and the implementation of sampling missions. Currently, the estimation of the surface particle sizes is mainly focusing on interpreting the thermal inertia with the infrared spectral data from ground-based or space telescopes, but this method show distinct errors compared with the imaging results of the orbiter. By analyzing some thermal infrared spectral data, a relationship between the particle sizes of the main rock-forming minerals (e.g. pyroxene, feldspar, olivine) and the slopes of their thermal infrared spectrum was found. Based on this relationship, a preliminary model for estimating the grain sizes (∼30–300 μm) of lunar or S-type asteroids' surfaces which are silicate minerals dominated was established, and the correlation coefficients (R2) for most of the rock-forming minerals were better than 90%. Six observational datasets of natural lunar and terrestrial samples are used to validate the model, and the results show a systematical overestimation of the ground-truth particle sizes, the potential causes are analyzed and an additional correction is used to eliminate the overestimation of the particle size prediction. These results are expected to provide guidance for interpretation of lunar and S-type asteroid surface sampling and spectral data.