Non-destructive detection of the quality attributes of fruits by visible-near infrared spectroscopy

How to rapidly and nondestructively detect fruit quality has always been a hot topic in fruit agriculture. In this paper, the quality detection of cream strawberry (CS) and rabbit-eye blueberry (RB) was studied by using visible-near infrared spectroscopy. The preprocessing methods such as de-trending, moving average smoothing, standard normal variable and baseline correction are used to reduce spectral data errors. The competitive adaptive reweighted sampling (CARS), successive projection algorithm (SPA) and the combination of the two algorithms (CARS+SPA) were used to reduce the data dimension. A partial least squares regression was used to establish the prediction model. The results showed that the characteristic wavelength extracted by CARS+SPA algorithm was the most suitable for predicting the contents of soluble solids content (SSC), total acid (TA), and vitamin C (VC) in CS and RB, and the determination coefficients were 0.96, 0.91, 0.91 and 0.93, 0.91, respectively. The relative percent deviation values of the prediction set were 4.47, 3.28, 3.69 and 3.74, 3.57, respectively. The correlation coefficients between the predicted value and the measured value of SSC, TA and VC content were 0.97, 0.92, 0.97 and 0.96, 0.95, respectively, which indicates that the established prediction model is very stable and reliable. This study can provide a theoretical basis for effectively solving the problem of rapid quality detection of multi-variety fruits and the development of multi-variety fruits quality detector.