Sugar Contents and Firmness of Apples Based on Multi-Spectral Imaging Technology

Abstract

The paper proposed a prediction method of apple sugar content and firmness based on multi-spectral imaging. Firstly, four characteristic wavelengths (670, 750, 780 and 810 nm) were selected by correlation coefficient method. The gray images of samples at different wavelengths were collected by multi-spectral imaging system, then fitted with Lorenz function, modified Lorenz function, Gaussian function and polynomial function, respectively. It was found that the fitting effect of modified Lorenz function was best. Therefore, the experiment was performed by multiple linear regression and partial least square regression analysis of sugar content and firmness with the fitting parameters of modified Lorenz function. The result showed that the prediction of multiple linear regression model was better than partial least squares regression model. The modeling correction correlation coefficient, calibration standard deviation, the prediction correlation coefficient and predicted standard deviation of sugar content were 0.8568, 0.6736, 0.8395 and 0.7068, respectively. The modeling correction correlation coefficient, calibration standard deviation, the prediction correlation coefficient and the predicted standard deviation of firmness were 0.8660, 0.3275, 0.8407 and 0.3555, respectively. The results also showed that this method was feasible for the prediction of apple sugar content and firmness.