SSC prediction of cherry tomatoes based on IRIV-CS-SVR model and near infrared reflectance spectroscopy

As one of the most important indexes of internal quality testing of fruit, soluble solids content (SSC) is significant for its rapid and efficient nondestructive testing by using near infrared reflectance spectroscopy (NIRS). In this article, 126 cherry tomatoes were selected as the research object. Reflectance spectra data of 228 bands in cherry tomatoes were acquired by the near infrared spectrometer and SSC was measured by the hand-held refractometer. Savitzky-Golay (SG) combined with multiplicative scatter correction (MSC) was used to preprocess the spectral data to reduce the effects of light scattering and other noise. Then, the dimensions of spectral data were reduced by iteratively retaining informative variables (IRIV) algorithm and 10 characteristic wavelengths were obtained, which were 1,080.37, 1,113.62, 1,117.3, 1,297.57, 1,301.02, 1,538.32, 1,540.40, 1,590.72, 1,615.94, and 1,636.89nm, respectively. Subsequently, support vector regression (SVR) and its two optimization models, PSO-SVR and CS-SVR, were respectively used to establish SSC prediction models based on full spectra and characteristic spectra. The experimental results showed the IRIV-CS-SVR model for SSC prediction achieved the accuracy with RP2 of 0.9718 and RC2 of 0.9845. Thus, it is feasible to use NIRS with IRIV-CS-SVR to make a rapid and efficient nondestructive SSC prediction of cherry tomatoes. Practical applicationsAs one of the important testing standards of fruit internal quality, SSC is of great significance for the rapid and efficient nondestructive testing. In this article, an iteratively retaining information variables (IRIV) algorithm is proposed to extract the characteristic wavelengths, and a regression model CS-SVR is established by combining the optimization algorithm cuckoo search (CS). This study shows that the model IRIV-CS-SVR has a certain effect on SSC prediction of cherry tomatoes.