Near-infrared (NIR) spectroscopy was used to measure trace chemicals, which could be useful for detection of pesticide residue in vegetable. Filter paper was used as the substrate. Pesticide solution was prepared by dissolving the commercial pesticide into distilled water at different concentrations. Samples were prepared by pipetting the solution onto the filter paper and then evaporated by vacuum drying oven. Then the spectra of samples were acquired in the region of 4000cm-1 to 10000cm-1 by NIR spectrometer. Partial least squares regression (PLSR) method and optimal band difference regression (OBDR) were used to establish prediction models respectively. Prediction results indicated that the PLSR models were able to predict the concentration of chlorpyrifos with 0.954 as the correlation coefficient of validation set, the OBDR models gave the best performance with 0.904 as the correlation coefficient of validation set. It could be concluded that NIR determination of pesticide was a low cost, environment friendly and potential method compare to the traditional methods. 