摘要
采用理化指标和电子鼻识别2种方法分别对6个成熟度(p1~p6)的果园荔枝进行识别。理化指标采样数据显示,荔枝果实直径、果核直径和果实净质量均随着果实的成熟而增大。p1—p4阶段,荔枝果皮绿色和黄色不断加深,亮度不断增大。p4—p6阶段,荔枝果皮亮度先增大后减小,颜色迅速变红,黄色成分先增加后减少。提取特征值后,采用主成分分析(PCA)、线性判别分析(LDA)、BP神经网络(BPNN)、简单相关分析(SCA)、典型相关分析(CCA)进行数据处理。理化指标识别法结合PCA和LDA对果园荔枝成熟度识别的正确率均为100%,能够较好地进行识别。但PCA识别结果中p1、p2和p3的距离较近,实际应用中易发生混淆。电子鼻识别法结合PCA和LDA分析均无法较好地对果园荔枝成熟度进行识别,电子鼻识别法结合BPNN对果园荔枝识别训练集的回判正确率为100%,测试集的识别正确率为92%,识别效果较好。SCA分析结果表明,在荔枝成熟过程中,除色差L*值外,其他各项理化指标均与电子鼻部分传感器的响应信号显著相关。CCA分析结果表明,电子鼻响应信号与理化指标整体相关性显著,电子鼻整体信号与部分理化指标相关性显著。证明了理化指标和电子鼻均能有效地识别水果品质信息变化,并为电子鼻替代理化指标识别法在水果品质信息监测上的应用提供了参考。
In order to explore the feasibility of using electronic nose substitute for physicochemical indexes to detect the quality information change of fruit,the physicochemical indexes identification method and electronic nose identification method were used for litchi samplings,which were in six different maturing stages( p1,p2,p3,p4,p5 and p6). The physicochemical indexes sampling results showed that fruit diameter, kernel diameter and fruit weight were increased as the fruit matured continuously. During stages of p1—p4,the green and yellow of fruits were continuously deepening,the brightness degree was continuously increasing. During stages of p4—p6,the brightness degree of fruit was increased at first and then decreased,the color was obviously gone red,the yellow was first deepened and then became shallow. After extracting the feature values,the principal component analysis( PCA),linear discriminant analysis( LDA),back propagation neural network( BPNN),simple correlation analysis( SCA) and canonical correlation analysis( CCA) were used for data process. Both results ofphysicochemical indexes identification method combined with PCA and LDA showed that litchi's maturing stage can be well identified,and both of their accuracies were 100%. But the distance between stages of p1,p2 and p3 were close when using PCA for analysis,which may be confused in practical classification and identification. However,litchi's maturing stage cannot be identified when using electronic nose combined with PCA or LDA for identification. When using electronic nose combined with BPNN for classification,the accuracies of train set and test set were 100% and 92%,respectively. SCA results showed that physicochemical indexes had significant correlation with electronic nose sensors' response except L*value during litchi's maturing process. CCA results showed that there was significant correlation between the whole physicochemical index set and the whole electronic nose sensors' response set. Part of physicochemical indexes had significant correlation with the whole electronic nose sensors' response set.The results proved the feasibility of using physicochemical index identification method and electronic nose identification method for detection of quality information change of fruit. It also provided reference for using electronic nose substitute for physicochemical indexes to detect the quality information change of fruit.
出处
《农业机械学报》
EI
CAS
CSCD
北大核心
2015年第12期226-232,共7页
Transactions of the Chinese Society for Agricultural Machinery
基金
国家自然科学基金资助项目(31571561)
现代农业产业技术体系建设专项资金资助项目(CARS-33-13)
广东省高等学校优秀青年教师培养计划资助项目(Y92014025)
关键词
荔枝
果园
电子鼻
理化指标
成熟度
相关分析
Litchi
Orchard
Electronic nose
Physicochemical indexes
Maturing stage
Correlation analysis