Abstract: The identification model of chlorpyrifos pesticide residues in wheat was established based on near infrared spectroscopy and stoichiometry. Wheat samples with different chlorpyrifos residues were prepared by spraying method. Near infrared diffuse reflectance spectra of samples were collected, and qualitative prediction models of factorization method, Euclidean distance method and quantitative prediction model of partial least squares method were established. The results showed that: The qualitative discrimination rate of factorization method was higher than that of Euclidean distance method, the characteristic bands of construction conditions were 5543.68~4643.72, 7582.43~6510.28 cm⁻¹, the preprocessing method was second derivative+vector normalization+5-point smoothing, the discrimination rate could reach 96%. The characteristic bands of the construction conditions of partial least squares quantitative discrimination model were 5543.68~4643.72 cm⁻¹, 7582.43~6510.28 cm⁻¹, the preprocessing method was to eliminate the constant offset, the main component fraction was 7, the cross validation coefficient (R²) was 99.46, and the mean square residual (RMSECV) was 1.15. The correlation between the actual value and the predicted value was good, and the minimum detection limit was 1 mg/kg, which could accurately determine the pesticide of Chlorpyrifos in wheat. The above results showed that the discrimination rate of the qualitative analysis model established by the factorization method was as high as 96%, which could realize the qualitative analysis and identification of chlorpyrifos pesticide residues in wheat, and the partial least squares method could realize the quantitative analysis and identification with good reproducibility.