基于高光谱的鸡蛋新鲜度检测

发布时间：2018-04-06 20:31

  本文选题：鸡蛋　切入点：新鲜度　出处：《光谱学与光谱分析》2016年08期

【摘要】：借助高光谱成像仪采集贮期白壳鸡蛋的透射高光谱数据,对比测量常规表征新鲜度的哈夫单位值,用Matrix Laboratory(MATLAB)和Statistical Analysis System(SAS)等软件,同时结合化学计量法对样品鸡蛋的高光谱数据进行分析处理,建立了基于高光谱技术的鸡蛋新鲜度预测模型。选用高光谱500~1000nm的波段作为敏感波段进行研究,用马氏距离剔除鸡蛋异常样本数据,并对鸡蛋高光谱数据进行了微分校正,通过比较发现高光谱二阶微分与鸡蛋哈夫单位值之间的线性度高,因此选用高光谱二阶微分数据来进一步研究,并对其进行了小波去噪、光滑处理及标准化处理。选用近年新提出来的competitive adaptive reweighted sampling(CARS)变量选取法对高光谱进行降维,提取出32个特征参数,建立了白壳蛋基于全波段的偏最小二乘法(partial least square,PLS)预测模型和基于特征参数的多元回归模型,验证集的相关系数分别为0.88,0.93,均方误差分别为7.565,6.44。用验证集的蛋对基于高光谱二阶微分全波段的偏最小二乘法预测模型、基于特征参数的多元回归模型分别进行验证,两个模型判别白壳蛋新鲜和不新鲜的最高准确率达100%,88%。
[Abstract]:The transmission hyperspectral data of white shell eggs during storage period were collected by means of hyperspectral imager, and Haff unit values, which were used to characterize freshness, were compared and measured. The software such as Matrix Laboratory MATLAB and Statistical Analysis system saps were used to measure the freshness of eggs.At the same time, the hyperspectral data of sample eggs were analyzed and processed by chemometrics, and a prediction model of egg freshness based on hyperspectral technology was established.Using the band of hyperspectral 500~1000nm as the sensitive band, the abnormal sample data of eggs are eliminated by Markov distance, and the hyperspectral data of eggs are corrected by differential correction.It is found that the linearity between the second order differential of hyperspectral and the unit value of egg Have is high, so the hyperspectral second order differential data is selected for further study, and wavelet denoising, smoothing and standardization are carried out.The hyperspectral dimensionality reduction is carried out by using competitive adaptive reweighted sampling method, which has been proposed in recent years, and 32 characteristic parameters are extracted. The prediction model and multivariate regression model of white shell egg based on partial least square method and feature parameter are established.The correlation coefficient of the verification set is 0.880.93, and the mean square error is 7.565 ~ 6.44.A partial least square prediction model based on hyperspectral second-order differential full-band method is used to predict the eggs of the verification set, and the multivariate regression model based on the characteristic parameters is verified respectively. The highest accuracy of the two models in judging the fresh and unfresh white shell eggs is 100% and 8888% respectively.
【作者单位】： 华中农业大学工学院;国家蛋品加工技术研发分中心华中农业大学;
【基金】：国家自然科学基金项目(31371771) 公益性行业(农业)科研专项(201303084) 国家科技支撑计划项目(2015BAD19B05)资助
【分类号】：O657.3;TS253.7

【相似文献】

中国期刊全文数据库 前1条

1 黄双萍;洪添胜;岳学军;吴伟斌;蔡坤;黎蕴玉;;基于高光谱的柑橘叶片磷含量估算模型实验[J];农业机械学报;2013年04期

中国硕士学位论文全文数据库 前2条

1 褚武道;基于PLS方法的铁观音茶树叶片营养成分含量高光谱估算[D];福建师范大学;2014年

2 何理;水稻叶片氮素含量及产量、相关品质高光谱预测模型的初步研究[D];扬州大学;2014年

，

本文编号：1718816