摘要
以红富士苹果为试材,在室温下贮存得到失重率0、5%、10%、15%和果心褐变果实,分别定义为1、2、3、4、5级新鲜度,采用LCR阻抗仪测定各等级共424个果实的14个生物阻抗参数在9个频率点下的126个特征值。通过稀疏主元分析-线性分类器(SPCA-LDC)模型试验,得出适宜的选样比例(训练样本数与测试样本数之比)为9∶1,主元数达到前39~45个时分级正确率达到最大值;构成至少39个主元的生物阻抗特征参数为44个;50次重复试验的分级正确率平均值为87.90%。
Bioimpedance properties of Fuji apple fruit with different freshness were investigated.Fruit with weight loss of 0,5%,10%,15% and core-browning under room temperature were chosen as samples corresponding to descending freshness grades of 1,2,3,4,5,respectively.126 observations of 14 bio-impedance parameters under nine frequencies of total 424 fruit from each grade were measured fruit by fruit with LCR impedance meter.Data were computed by using the sparse principal component analysis-linear classifier(SPCA-LDC) model.Results indicated that classification accuracy reached to maximum stably when the ratio of training sample to test sample of 9∶1 and former 39-45 principal component(PC) were adopted.The least 39 PCs were consisted of 44 from the 126 observations.Average rate of classification accuracy reached to 87.90% within 50 test repeats.
出处
《农业机械学报》
EI
CAS
CSCD
北大核心
2013年第2期147-152,共6页
Transactions of the Chinese Society for Agricultural Machinery
基金
国家自然科学基金资助项目(61202188)
农业部现代农业产业技术体系建设专项资助项目
陕西省自然科学基金资助项目(2010JQ8019)
关键词
苹果
新鲜度
生物阻抗
无损测定
稀疏主元分析
线性分类器
Apple
Freshness
Bio-impedance
Non-destructive detection
Sparse principal component analysis
Linear classifier