Abstract:In order to improve the accuracy and robust of NIR spectroscopy modules in predicting the firmness of strawberry,simulated annealing algorithm (SAA) was used to select the wavenumbers in NIR spectra. A preprocessing method was also selected to adapt the SAA. Firstly, 150 strawberries were selected to collect NIR spectra. Secondly, preprocessing methods, such as SNV, MSC, 1st order derivation, 2nd order derivation, were used to denoise the NIR spectra of strawberry. Thirdly, 24 wavenumbers were selected by simulated annealing algorithm. At last, partial least square was employed to establish the calibration models of firmness. The calibration model was obtained with the correlation coefficient rc of 0.9342, the root mean square error of calibration of 0.665N/cm2 and the correlation coefficient rp of 0.9197,the root mean square error of prediction of 0.673N/cm2. The results show that SAA can improve the robust and accuracy and simplify NIR spectra models.