Models establishment of watermelon qualities based on electronic tongue analysis
In order to explore the E-tongue technology for rapid detection of internal quality of watermelon,the taste com-ponents of 200 watermelons were examined using the E-tongue technology with L600 watermelon as the test material,sol-uble solids,titratable acid,moisture,vitamin C and total phenol content,sugar-acid ratio and pH were determined by tradi-tional detection methods,and partial least squares,random forest,support vector machine and K-nearest neighbour algo-rithms were used to establish quality prediction model.The results showed that soluble solids content,sugar-acid ratio,vi-tamin C content and total phenol content of watermelon tended to increase,pH tended to increase and then decrease,and titratable acid and water content tended to decrease during the growth process.In the model prediction results,the model prediction of the qualities of watermelon soluble solids content,sugar-acid ratio,pH and moisture content were better than titratable acid,vitamin C and total phenol content.The Rp2 of the random forest algorithm for the prediction of watermelon soluble solids content,titratable acid content,sugar-acid ratio and moisture content were 0.884,0.798,0.891 and 0.875,which were 16.9％,19.1％,21.6％and 13.6％higher than those of the partial least square algorithm,respectively;and the RP2 of the partial least square algorithm for pH was 0.881,which was 31.7％higher than that of the support vector ma-chine algorithm;and the K-nearest neighbour algorithm predicted vitamin C and total phenol content with Rp2 of 0.731 and 0.753,which were 1.67％and 24.5％higher than the random forest algorithm,respectively.The results indicated that it is feasible to apply the electronic tongue technique to predict the internal quality of watermelon.
万方数据
维普
