Pea (Pisum sativum L.) is an important edible legume. Feasibility of the Fourier Transform Near-Infrared Spectroscopy (FT-NIRS) on estimating quality traits in pea was evaluated in current study. A total of 190 pea accessions were involved with their protein, starch, oil and total polyphenol content chemically analyzed. After obtaining spectra of the samples in milled powder and intact seed forms, partial least squares (PLS) regression was applied for model development. Models of powder were generally superior to models in intact seed. The optimal models were powder-based for protein and starch with residual predictive deviation (RPD) of 5.88 and 5.82 as well as coefficients of correlation (r2) of 0.99 and 0.99, respectively. High values of correlation coefficient (r2) revealed that models had good predictive capacities for rapid germplasm analysis of pea. To explore the relationship between quality traits and producing regions, 150 pea varieties with specific information were analyzed by two-step cluster analysis. Three distinct groupings were obtained with obvious features. Group1 was in low protein content. Group2 was in high total polyphenol content. Group 3 was in high protein, starch and oil content. The nutrition contents were affected by longitude, latitude and altitude of planting location as well as seeding date. These results can provide reliable information for the collection of excellent germplasm resource in good quality traits.