高光谱图像技术定量检测香醋醋醅水分分布均匀性  

Quantitative Detection of Homogeneity of Moisture Content Distribution in Vinegar Culture by Hyperspectral Imaging Technique

作　　者：石吉勇[1] 胡雪桃 朱瑶迪[1] 邹小波[1] 陈武[1] 周佐花 

机构地区：[1]江苏大学食品与生物工程学院,江苏镇江212013

出　　处：《中国食品学报》2018年第2期250-255,共6页Journal of Chinese Institute Of Food Science and Technology

基　　金：国家自然科学基金项目(31772073);江苏省重点研发项目(BE2016306);江苏省国际合作项目(BZ2016013);江苏省六大人才高峰项目(GDZB-016)

摘　　要：水分含量是影响镇江香醋醋醅发酵的重要因素,探索醋醅中水分分布均匀性的定量描述方法,可为精确调控香醋发酵工艺提供科学依据。利用高光谱图像技术检测醋醅中水分含量的二维分布图,以水分分布图的标准差作为定量指标,首次提出醋醅水分分布均匀性的定量描述方法。首先利用醋醅高光谱图像信息结合遗传算法建立水分含量检测模型;其次提取各像素点对应的高光谱图像信息代入水分含量检测模型,快速计算出所有像素点对应的水分含量,从而描绘醋醅水分含量二维分布图;最后计算水分含量分布图的标准差,以此定量表征各像素点水分含量与所有像素点水分含量均值之间的离散程度,两者之间离散程度越小,水分分布越均匀。研究表明:高光谱图像技术定量分析醋醅水分分布均匀性是可行的,为定量描述固态发酵关键指标的分布均匀性提供了新的检测手段。The moisture content plays a critical role in Zhenjiang vinegar culture. The method of quantitatively describing homogeneity of moisture distribution in vinegar culture can be a scientific basis for accurately regulating the process of Zhenjiang vinegar fermentation. In this work, hyperspectral imaging was employed to measure the two-dimensional map of moisture distribution in vinegar culture. The method for quantitatively describing homogeneity of moisture, firstly,was proposed using standard deviation of moisture distribution map as quantitative indicator. The hyperspectral images of vinegar culture were obtained by hyperspectral camera in the wavelengths of 431~963 nm. The 81 spectra data of vinegar culture were extracted from the regions of interest in the Hyperspectral image. GA-PLS was the optimal model selected from the five models（PLS, i PLS, si PLS, bi PLS, GA-PLS）. With twelve wavelengths（544, 546, 531, 535, 575,832, 835, 840, 849, 895, 896, 901 nm） optimized by GA among 618 wavelengths, 0.877 of Rpand 0.363 of RMSEP were acquired in GA-PLS model indicating the relationship between the spectrum and moisture content. The concentration of moisture at each pixel of hyperspectral images were calculated by inputting the hyperspectral images with twelve wavelengths into the established GA-PLS model. The distribution of moisture content was gotten via the spatial position of each pixel. Finally, standard deviation of moisture content obtained from the moisture content of each pixel was implemented as a measure of the dispersal extent between moisture content of each pixel and average moisture content of all pixels. The smaller the degree of dispersion between those two is, the more uniform the moisture content distribution will be. The results indicate that it is feasible for hyperspectral image to quantitatively describe homogeneity of moisture distribution in vinegar culture, suggesting that hyperspectral image provides a new method for detecting homogeneity of crucial indicators in solid-state ferme

关 键 词：醋醅 水分分布均匀性 高光谱图像技术 遗传算法 偏最小二乘法 

分 类 号：O657.3[理学—分析化学] TS264.22[理学—化学]