Meconopsis quintuplinervia Regel,a traditional Tibetan medicine,is usually used in the part of flower powder as medicine.It has been reported that this medicinal species can be used to treat pneumonia,hepatitis,etc.while different parts of its powder presented different therapeutic effect.Therefore,in this study,the different parts of powdered flower,leaf and whole grass in this medicine were considered as the research objects so as to solve the problem of discrimination of different medicinal part.First,mid-infrared spectra of different parts in M. quintuplinervia were collected,and spectral features and one-way ANOVA of absorbance values were also analyzed among different parts.It was found that it was difficult to discriminate parts in visual although the MIR spectra features varied in each part.Then,a statistical analysis revealed that the difference of absorbance values among different peaks was very significant.The results indicated that the establishment of a part discriminant model was available. Finally,orthogonal test and range analysis were also processed to optimize the modeling conditions with recognition rate and prediction rate as indicators.The results showed that the optimized modeling method was Distance Match+SNV+second derivative spectrum (D2),and the modeling wavenumber were set in two different spectral ranges 3 031-2 810 cm^-1 and 1 800-1 450 cm^-1,respectively.The recognition rate of model was 99.05%,and the prediction rate was 96.19%.Then the recognition rate of test set of 96.67% was obtained and performed a well model.Meanwhile,additional recognition performance was evaluated in order to judge the effect of the model.The results showed a lower ER(error rate) values than 0.02.and the value of TPR(true positive rate),TNR(true negative rate) and F1 were between 0.97 and 1.00.Therefore,it was revealed that the established model in our study can rapidly and effectively identify the different parts of M. quintuplinervia.