Near-infrared spectroscopy can reflect the internal hydrogen-containing chemical bond stretching vibration and ensemble frequency absorption information of the sample, with the advantages of high-speed, economical, reproducible and environmentally friendly analysis, commonly used in food, pharmaceuticals and materials detection for analysis. Peas are one of the most important cultivated crops in the world, widely grown and distributed, with nutritional properties such as high starch, high protein as well as low lipid content, which consumers have long loved. In order to clarify the modeling differences in the NIR spectra of peas from different origins, modeling analysis was performed on peas from different origins. In this study, 42 pea samples collected from different regions of Nanyang city, Henan province, were investigated. Firstly, pea samoles’ nutritional components (total starch, protein, moisture, ash, and lipid) were determined. Then, with an emphasis on using the integrating sphere diffuse reflectance technology in Near-infrared spectroscopy, the spectra of different pea samples were collected in the wavelength range of 12 000~4 000 cm-1. By combining different pre-processing methods with discriminant analysis models (DA) to obtain the optimal pre-processed data and combining principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA) along with orthogonal partial least squares discriminant analysis (OPLS-DA), the differences in the spectral characteristics were screened and analyzed, constructing and verifying the identification models of Nanyang peas.The results show that the overall difference in nutritional composition and content of Nanyang peas in different regions is slight (total starch 36.30%~46.93%, protein 16.37%~25.50%, moisture 6.78%~9.16%, ash 2.29%~3.38%, lipid 0.37%~1.43%), and the results of the discriminant model established based on Near-infrared spectroscopy showed that the accuracy of discriminant analysis for the DA model could reach 92.4%. The predictive abilities obtained from PCA, PLS-DA and OPLS-DA models were 96.7%, 85.1% and 83.6%, respectively, indicating that the above models can achieve accurate classification and identification of geographical origin for Nanyang peas. In addition, the results of the different bands among different geographical origins screened by the variable importance projection (VIP>1.0) method showed that 4 710~4 000, 5 320~5 200 and 7 200~6 220 cm-1 could be used as specific detection bands to identify geographical origins of Nanyang peas. Therefore, this study can provide a methodological basis to construct a database for the identification and traceability of pea’s origin in different regions.