河南农业科学
河南農業科學
하남농업과학
JOURNAL OF HENAN AGRICULTURAL SCIENCES
2014年
1期
149-153
,共5页
吴瑞梅%赵杰文%陈全胜%黄星奕
吳瑞梅%趙傑文%陳全勝%黃星奕
오서매%조걸문%진전성%황성혁
绿茶%茶汤汤色%感官评价%特征提取%BP神经网络法
綠茶%茶湯湯色%感官評價%特徵提取%BP神經網絡法
록다%다탕탕색%감관평개%특정제취%BP신경망락법
green tea%tea infusion color%sensory evaluation%feature selection%BP-ANN
为弥补感官审评法评定绿茶汤色品质的不足,研究了绿茶汤色品质感官评价的色差仪表征方法。采用色差仪获取茶汤的色度值,分别利用逐步回归法和主成分分析(PCA)法提取汤色特征变量,建立汤色特征变量与其感官审评分值之间的相关模型。结果表明,利用PCA方法提取特征变量的效果好于利用逐步回归法。基于主成分分析法提取的特征变量,分别建立特征变量与感官评分之间的PLS和BPANN模型对绿茶汤色品质进行预测,其准确度可靠,但BPANN模型性能好于PLS模型。用BPANN模型对预测集样本进行测试时,预测集中各样本的预测值与实测值间相关系数为0.816,预测均方根误差为2.505。可见,利用色差仪结合BPANN方法能有效表征绿茶汤色品质。
為瀰補感官審評法評定綠茶湯色品質的不足,研究瞭綠茶湯色品質感官評價的色差儀錶徵方法。採用色差儀穫取茶湯的色度值,分彆利用逐步迴歸法和主成分分析(PCA)法提取湯色特徵變量,建立湯色特徵變量與其感官審評分值之間的相關模型。結果錶明,利用PCA方法提取特徵變量的效果好于利用逐步迴歸法。基于主成分分析法提取的特徵變量,分彆建立特徵變量與感官評分之間的PLS和BPANN模型對綠茶湯色品質進行預測,其準確度可靠,但BPANN模型性能好于PLS模型。用BPANN模型對預測集樣本進行測試時,預測集中各樣本的預測值與實測值間相關繫數為0.816,預測均方根誤差為2.505。可見,利用色差儀結閤BPANN方法能有效錶徵綠茶湯色品質。
위미보감관심평법평정록다탕색품질적불족,연구료록다탕색품질감관평개적색차의표정방법。채용색차의획취다탕적색도치,분별이용축보회귀법화주성분분석(PCA)법제취탕색특정변량,건립탕색특정변량여기감관심평분치지간적상관모형。결과표명,이용PCA방법제취특정변량적효과호우이용축보회귀법。기우주성분분석법제취적특정변량,분별건립특정변량여감관평분지간적PLS화BPANN모형대록다탕색품질진행예측,기준학도가고,단BPANN모형성능호우PLS모형。용BPANN모형대예측집양본진행측시시,예측집중각양본적예측치여실측치간상관계수위0.816,예측균방근오차위2.505。가견,이용색차의결합BPANN방법능유효표정록다탕색품질。
To make up for the defect of sensory evaluation method for evaluating green tea infu-sion color quality ,the quantification methods for the sensory evaluation of green tea infusion color based on the color difference meter were studied in this study .A color difference meter was used to get the colorimetric values of tea infusion for each sample .Stepwise regression and principal component analysis (PCA ) methods were respectively used to extract the colorimetric features from the colorimetric values of tea infusion .The results showed that the model based on the fea-ture variables selected by PCA got better result .Then partial least squares(PLS) regression and back-propagation artificial neural network(BP-ANN) methods were performed to build the cali-bration models based on the relationship between sensory scores evaluated by tea tasters for the infusion color of tea samples and the feature variables extracted by PCA ,respectively .The results indicated the predicted results by PLS and BP-ANN models were consistent with the evaluation results of the sensory profiling panels in green tea infusion color study ,but the performance of BP-ANN model was better than that of PLS .The correlation coefficient and root mean square er-ror of prediction of BP-ANN model were 0 .816 and 2 .505 in the prediction set ,respectively .The study demonstrates that a color difference meter combined with BP-ANN method can quantify the quality of infusion color for green tea .