Abstract: The internal qualities of apple were detected with spectral imaging in this study. The spectral imaging in wavelength of 632, 650, 670, 780, 850 and 900 nm were respectively captured. The functions of lorentzian distribution (LD), gaussian distribution (GD) and exponential distribution (ED) with three parameters were used to fit scattering profiles for all wavelengths. LD was found to be the best function for fitting gray distribution of imaging. The multi-linear regression model related with lorentzian parameters to fruit firmness were developed with best single wavelength, double wavelengths, three wavelengths and four wavelengths respectively. The best model with respect to four wavelengths was able to predict apple firmness with r=0.880. Results showed that the multi-spectral scattering imaging provides a nondestructive means for measuring apple firmness, and is a fast and easy method to perform.

Key words: spectral imaging, firmness, apple, multi-linear regression