玉米汁辨识中的传感器阵列优化  

Optimization of sensor array for identification of corn juice

作　　者：刘晶晶[1] 孙永海[1] 丁健峰[1] 孙钟雷[2] 

机构地区：[1]吉林大学生物与农业工程学院,长春130022 [2]长江师范学院生命科学与技术学院,重庆408100

出　　处：《吉林大学学报（工学版）》2013年第2期538-543,共6页Journal of Jilin University:Engineering and Technology Edition

基　　金：国家自然科学基金项目(31271861);'863'国家高技术研究发展计划项目(2008AA100802)

摘　　要：针对玉米汁饮料品质评价自动化关键技术问题,对用于玉米汁饮料分类辨识的传感器阵列进行了筛选优化。采用因子分析对各传感器响应信号之间的内在关联进行了研究,依据相关系数和旋转后的因子含义将传感器分为6类,并组合成12组不同的传感器阵列。利用概率神经网络判定界面可以任意复杂的特点,选择平滑参数为0.01时构建的判定界面表征出不同组传感器阵列所得数据对分类结果的影响差异。选择S4,S2,P2,P3,S6,S7传感器组合,构建新的传感器阵列,并对7种玉米汁饮料进行辨识。优化后的传感器阵列对玉米汁饮料的分类准确率为98.016%,高于未经优化的传感器阵列的分类准确率(94.444%)。The sensor array for identification of corn juices was optimized for the key technical issues of automated quality evaluation of corn juices. The inherent relationship among the response signals of sensors was analyzed by the factor analysis, and the sensors of the array were divided into 6 categories according to the correlation coefficients and the component plot in rotated space, and combined into 12 different sets of sensor array. Because the decision boundary surface of the probabilistic neural network can be arbitrary complex, the decision boundary surface buitt at selected smoothing parameter 0. 01 characterizes the impact differences of the data acquired by different sets of sensor array on the identification results. Selecting the sensors named S4, S2, P2,P3, S6 and S7 to build a new sensor array to identify 7 kinds of corn juices. The identification accuracy rate for corn juices by the optimized sensor array was 98. 016%, which was higher than 94. 444% by the non-optimized one.

关 键 词：食品科学技术 传感器阵列优化 概率神经网络 因子分析 玉米汁分类辨识 

分 类 号：TS207.3[轻工技术与工程—食品科学]