Abstract: Remote sensing is a crucial technique for identifying the composition and physical state of materials on celestial surfaces. Among various methods, mid‑infrared (MIR) spectroscopy offers significant diagnostic capability for determing mineral crystal structures, chemical composition, and physical states. Spectral indicators such as the Christiansen Feature (CF), Reststrahlen Bands (RBs), and Transparency Feature (TF) in MIR reflectance spectra collectively form a spectroscopic framework for identifying mineral composition and physical properties like particle size. Currently, laboratory measurements of MIR reflectance spectra primarily include bi-conical diffuse reflectance, directional‑hemispherical integrated reflectance, and micro‑reflectance techniques. However, constrained by experimental conditions and sample states, existing models for mineral identification and particle‑size prediction are often built on data acquired through different measurement methods, which may lead to systematic shifts in spectral features and variations in particle‑size effects, thereby limiting the extrapolation applicability of such models. To address this issue, this study elected typical lunar surface minerals—olivine, pyroxene, and plagioclase—prepared as samples across five characteristic particle‑size ranges (<26 μm, 26~43 μm, 43~74 μm, 74~125 μm, 125~200 μm), and performed spectroscopic measurements using micro‑reflectance on signle particle, bi-conical diffuse reflectance and directional‑hemispherical reflectance on powdered samples. By extracting and comparing six key MIR spectral parameters, we quantitatively revealed the systematic offsets and particle‑size dependencies of these parameters across different measurement methods. Futhermore, it clarified the respective applicability and data interoperability conditions of micro‑reflectnace, bi-conical diffuse reflectance, and directional‑hemispherical reflectance methods for particle‑size prediction and mineral identification. These findings provide essential methodological support for the construction and interpretation of remote‑sensing spectral models.