基于高光谱图像的香醋醋醅均匀性快速判别  

Quick Discrimination of Uniformity Degree of Zhenjiang Balsamic Vinegar Grains Based on Hyperspectral Imaging Technology

作　　者：邹小波[1] 申婷婷[1] 朱瑶迪[1] 石吉勇[1] 胡雪桃 沈冰[1] 

机构地区：[1]江苏大学食品与生物工程学院,镇江212013

出　　处：《农业机械学报》2016年第1期215-220,共6页Transactions of the Chinese Society for Agricultural Machinery

基　　金：国家高技术研究发展计划(863计划)项目(2011AA100807);国家自然科学基金项目(61301239);全国优秀博士基金项目(200968);新世纪优秀人才项目(NCET-11-00986);江苏省杰出青年基金项目(BK20130010);江苏省研究生创新基金项目(KYLX_1070)

摘　　要：以镇江香醋固态发酵过程中总酸质量分数和p H值为表征指标,利用高光谱图像技术和化学计量学实现翻醅均匀性快速判别。以翻醅前后醋醅为研究对象,首先采集其高光谱信息,同时检测对应的总酸质量分数和p H值;然后以联合区间偏最小二乘（si PLS）和遗传算法（GA）优选特征变量;采用偏最小二乘（PLS）和最小二乘支持向量机（LS-SVM）方法建立表征醋醅均匀性指标的快速预测模型;最后使用最优模型计算高光谱图像每个像素点的总酸质量分数和p H值,伪彩色处理,获得醋醅总酸质量分数和p H值分布图。结果表明,总酸质量分数和p H值最佳模型的相关系数（R）分别为0.775 1和0.797 4、预测均方根误差（RMSECP）为0.389%和0.041 7;分布图中翻醅前总酸质量分数和p H值范围为3.0%-7.8%和3.5-4.2,醋醅局部有明显结块,分布不均匀,翻醅后总酸质量分数和p H值范围较集中,为4.8%-7.0%和3.6-3.9,醋醅分布较均匀。结果表明利用高光谱技术快速预测醋醅均匀性是可行的。This study selected total acid content and p H value as characterization indicators and used hyperspectral imaging technology and chemometric methods to discriminate uniformity of Zhenjiang balsamic vinegar grains. First,hyperspectral transmission images in 432 - 960 nm of vinegar grains,total acid content and p H value were acquired. Secondly,PLS and LS-SVM method were used to establish uniformity indicators content prediction models after preferred variables which were selected by GA and si PLS. The root mean square error of prediction（ RMSEP） and correlation coefficient（ R） for the test set are 0. 389% and 0. 775 1 for total acid content,0. 041 7 and 0. 797 4 for p H value. Total acid content and p H value in each pixel point on the vinegar grains were obtained by the above prediction models.After pseudo-color processing,the distribution map of total acid content content and p H value before and after turning up the vinegar grains were obtained. In the distribution maps,the range of total acid content and p H value in the grain were 3. 0% to 7. 8%,3. 5 to 4. 2 before the overturning,and 4. 8% - 7. 0%,3. 6 - 3. 9 after the overturning. By overturning the grains,the degree of uniformity was improved while high local concentrations phenomenon still exists. The overall results revealed that hyperspectral imaging technology is a promising technique to discriminate the degree of uniformity of grains after overturning rapidly and nondestructively.

关 键 词：镇江香醋 固态发酵 醋醅分布 高光谱图像 均匀性 

分 类 号：TS264.22[轻工技术与工程—发酵工程] TP391.41[轻工技术与工程—食品科学与工程]