Abstract: In this paper, the erucic acid content of rapeseed oil and the correction model of near-infrared transmission spectrum were established based on partial least square(PLS) method by using multichannel PDA near-infrared spectroscopy instrument. The influences of window size of polynomial derivation and smoothing and wavelength selected by correlation coefficient on the correction model were discussed. The predicted precisions of ten samples were analyzed by the index of predicted relative coefficient(R_P) and root-mean-square error(RMSEP). The result showed that the first derivation with seven points and smooth pretreatment is the best while adopting PLS modeling on the whole spectra, and the results of R_P and RMSEP are 0.379 and 1.659, respectively. When selecting wavelength by relative coefficient, R_P and RMSEP are 0.958 and 0.963, respectively. Compared with the results of the two methods, the latter R_P improves by 29% and RMSEP reduces by 42%. Therefore, the conclusions can be drawn that combining polynomial derivation and smooth pretreatment with correlation coefficients will improve the steadiness and predicted precisions of correction model. It shows that it is feasible to quickly determine the erucic acid content in rapeseed oil by multichannel near-infrared spectroscopy.