Abstract: In order to establish mathematical model between near infrared diffuse reflection （NIR） spectroscopy and hardness of blueberry， the hardness of TPA and puncture was detected by texture analyzer. After scanning the blueberry in the spectral region between 400-2 500 nm， the influence of different spectral preprocessing methods on the model was compared. Then 40 unknown samples were predicted by the optimal model， in order to verify the veracity and stability of the model. The results showed that the correlation coefficient of prediction（R2p）of TPA hardness was 0875 2， the root mean square error of prediction （SEP） was 0251 and relative percent deviation of prediction （RPD） was 282; correlation coefficient of prediction（R2p）of puncture hardness was 0717 8， the root mean square error of prediction （SEP） was 0024 and relative percent deviation of prediction （RPD） was 210. Therefore the model was stable， which could meet the requirements of practical application and indicated that the fast nondestructive measurement of blueberry hardness using NIR spectroscopy technique was feasible.

Key words: blueberry, near infrared diffused spectroscopy, TPA hardness, puncture hardness, non\, destructive detection