Research on Internal Quality Detection Method of Cherry Tomatoes Based on Improved WOA-LSSVM

Based on near-infrared spectroscopy and statistical methods, a rapid and non-destructive testing method for the internal quality of cherry tomatoes was proposed. First, the near-infrared spectrum of the sample was collected, and five preprocessing methods, Multiplicative Scatter Correction, Savitzky-Golay convolution smoothing, Savitzky-Golay convolution first derivative, De-trending, Standard Normal Variate, and SNV were used to eliminate spectral interference and screen out the best preprocessing method; then use the Successive Projections Algorithm, Stability Competitive Adaptive Reweighted Sampling, Genetic Algorithm, and the introduction of automatic ordered predictor selection algorithm for Improved Genetic Algorithm Four characteristic wavelength extraction methods reduce variable redundancy and select the optimal characteristic wavelength extraction. Method; finally, combined regression method-combining von Neumann topology, roulette selection, tournament selection and adaptive weights with whale algorithm to improve the algorithm, using the Improved Whale Optimization Algorithm, and based on Particle Swarm Optimization-BP Neural Network was compared with the Whale Optimization Algorithm-Least Squares Support Vector Machine, and the prediction models for the internal quality content of cherry tomatoes were established respectively. The results showed that the De-trending-IGA-IWOA-LSSVM model was used for the best soluble solid content in the internal quality of cherry tomatoes, where the coefficient of determination of the calibration set and prediction set were 0.917 2 and 0.866 7, respectively, the corrected root mean square error and the predicted mean square The root error was 0.542 3 and 0.768 2, and the relative error of prediction reached 2.592 9; the SG-IGA-IWOA-LSSVM model was used to predict the Vitamins C content the most accurate, and the coefficient of determination of the calibration set and prediction set were 0.857 6 and 0.821 6, respectively, and the corrected root mean square The error and prediction root mean square error are 0.661 4 and 0.634 2, respectively, and the prediction relative error reaches 2.078 5. The above results show that the combination of near-infrared spectroscopy and statistical methods can achieve rapid and non-destructive prediction and analysis of the internal quality of cherry tomatoes.