Abstract: A rapid method based on near infrared diffuse reflectance spectroscopy was established for the detection of endosulfan residue on tomato pericarp. A Boosting partial least square (Boosting-PLS) regression was applied for building the quantitative models with second derivatives (polynomial order=2，width of the window=17 points) and Rubberband baseline correction (n=64) as the pre-processing method. Promising results were achieved with determination coefficent of 0.992 and the root mean square error of cross validation/prediction (RMSECV and RMSEP) of 2.82/2.79. The confidence interval(α=0.05) of average recoveries for calibration and prediction sets were (101.1±1.4)% and (98.9±2.6)%，respectively. The method showed the potential of being a rapid，economical and environmentally acceptable method for detecting endosulfan residue on fruit pericarp.