【摘要】 本研究构建了一个快速准确的玫瑰香葡萄货架期品质判别预测模型，以最大化玫瑰香葡萄的商品价值。通过利用电子鼻分析技术对不同温度(0、10和20℃)条件下玫瑰香葡萄货架期内挥发性气体进行检测，采用偏最小二乘法(Partial Least Squares Regressions, PLS)和BP神经网络(Back Propagation Artificial Neural Network, BP-ANN)对货架期内玫瑰香葡萄的可溶性固形物(Solid Soluble Content, SSC)和总酸(Total Acid, TA)建立预测模型；为提高新鲜度判别模型的准确性，通过系统聚类分析将SSC、TA与感官评价信息进行融合，结合遗传算法优化支持向量机构建玫瑰香葡萄货架期新鲜度判别模型。结果表明：PLS和BP-ANN模型均可有效预测SSC和TA的含量，其中BP-ANN模型的预测精度更高(SSC模型的R~2=0.969 4,RMSE=0.009 4;TA模型的R~2=0.918 3,RMSE=0.002 5);基于品质信息融合的玫瑰香葡萄新鲜度判别模型的判别准确率为95%,本研究为更准确的预测玫瑰香葡萄的理化指标和判别新鲜度提供新的思路。更多还原

【Abstract】 This study constructed a fast and accurate prediction model of quality discrimination to ensure consumer consumption safety and maximize the commodity value of rose grape. The volatile gases of different temperatures(0, 10 and 20 ℃), Using the partial least-squares regressions(PLS) and the back-propagation artificial neural network(BP-ANN) for solid soluble content(SSC) and total acid(TA) Establish a prediction model; To improve the accuracy of the freshness-discriminative model, Fusing SSC, TA and sensory evaluation information by systematic clustering analysis, Optimize the freshness of rose grape. The results show that both PLS and BP-ANN models can effectively predict the content of SSC and TA, in which the prediction accuracy of BP-ANN model is higher(SSC: R~2=0.969 4, RMSE=0.009 4; TA: R~2=0.918 3, RMSE=0.002 5); the discrimination accuracy of the model based on quality information fusion is 95%, which provides new ideas for more accurate physical and chemical indexes and freshness.更多还原