Calibration model for firmness of Agaricus bisporus during storage was developed based on near-infrared (NIR) diffuse reflectance spectroscopy quantitative analysis techniques. Partial least squares (PLS) regression was carried out to analyze the spectroscopy. Four different spectral pretreatment methods (first derivative, second derivative, SNV, SMC) were used to compare calibration results for firmness. The result showed that the best calibration models，in the selected wavelength range of 5000~10000cm-1，could be obtained by the second derivative spectrum with the determination coefficient of calibration of 0.9471, and the determination coefficient of prediction of 0.8261. The results indicated that the detection method based on near-infrared diffuse reflectance spectroscopy was simple, and could be used for non-destructively evaluating the changes of firmness in Agaricus bisporus during storage in time.