Sichuan Ophiopogon japonicus is an important medicine and is widely used in clinic as it has the effects of nourishing yin and nourishing fluid，moistening lung and clearing heart．However，time-consumption and no on-site detection of the original medicinal material have always been the urgent problems to solve in its quality analysis．In this paper，near infrared spectroscopy was applied to the quality analysis of Sichuan Ophiopogon japonicus．Different grades of Sichuan Ophiopogon japonicus were collected，and the content data of water，extract and total saponins were determined by segmented experiment．In-situ and non-destructive detection of Sichuan Ophiopogon japonicus granules was performed by diffuse reflectance mode，and its near infrared spectra were collected．Meanwhile，different spectral pretreatment methods were used to deal with the spectrum in order to reduce the interference caused by particle size to the spectrum．The best pretreatment methods for moisture，extract and total saponins were multivariate scattering correction combined with standardization，multivariate scattering correction combined with first-order derivative，SG smoothing and mean center，respectively．Moreover，the variable screening method was used to extract the effective information in the spectrum，the best band screening method for moisture was piecewise random frog，and the best band selection method for extract and total saponins was variable importance projection．Finally，the model was established according to the optimal pretreatment method and the optimal variable screening method．The root mean square errors（RMSEP） of the model for moisture，extract and total saponins were 0.165 5%，0.401 9% and 0.078 4%，respectively，and the determination coefficients for validation（,R,${\mathrm{v}\mathrm{a}\mathrm{l}\mathrm{2}}_{}^{}$,https://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=49026948&type=,https://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=49026943&type=,2.53999996,4.31799984,） were 0.965 1，0.696 5 and 0.803 6，respectively．The relative percent deviations（RPDs） were 4.50，2.68 and 2.22，respectively．The RPDs for the three quality indicators were all greater than 2，indicating that the performance of the model was good，and the model built in this paper could meet the requirements for quality analysis．In this paper，a near-infrared quantitative model for water，extract and total saponins in Sichuan Ophiopogon japonicus was established by in-situ detection and parameter optimization，which could meet the needs of “rapid and non-destructive” in collection，processing and product testing，realizing a new real-time and green analysis for the quality control of Sichuan Ophiopogon japonicus，and providing a reference for the quality analysis of Sichuan Ophiopogon japonicus.