融合密度与光谱特征的苹果霉心病无损检测  

Research on non-destructive detection method of moldy apple core by fusing density and spectral features

作　　者：张佐经[1,2,3] 付新阳 陈柯铭 赵遵龙 张仲雄 赵娟 ZHANG Zuojing;FU Xinyang;CHEN Keming;ZHAO Zunlong;ZHANG Zhongxiong;ZHAO Juan(College of Mechanical and Electronic Engineering,Northwest A&F University,Yangling 712100,China;Key Laboratory of Agricultural Internet of Things,Ministry of Agriculture and Rural Affairs,Yangling 712100,China;Shaanxi Key Laboratory of Agricultural Information Perception and Intelligent Service,Yangling 712100,China)

机构地区：[1]西北农林科技大学机械与电子工程学院,陕西杨凌712100 [2]农业农村部农业物联网重点实验室,陕西杨凌712100 [3]陕西省农业信息感知与智能服务重点实验室,陕西杨凌712100

出　　处：《食品与发酵工业》2022年第15期281-287,共7页Food and Fermentation Industries

基　　金：国家自然科学基金项目(31701664)。

摘　　要：针对近红外漫反射光谱对苹果霉心病判别准确率较低的问题,提出了一种融合密度特征与漫反射光谱的苹果霉心病多因子无损检测方法。基于光谱采集平台获取195个富士苹果的漫反射光谱(200~1100 nm)信息,利用WLD-600密度仪获取苹果密度信息,采用标准正态变量变换(standard normal variable transformation,SNV)对光谱数据进行预处理,竞争性自适应重加权采样法(competitive adaptive reweighted sampling,CARS)和连续投影算法(successive projection algorithm,SPA)结合用于提取与霉心病相关的特征光谱,分别以密度、特征光谱、密度+特征光谱作为模型因子,建立偏最小二乘判别(partial least squares discriminant analysis,PLS-DA)、Fisher判别、支持向量机(support vector machine,SVM)和最小二乘支持向量机(least squares support vector machine,LS-SVM)4种不同的苹果霉心病判别模型。结果表明,在不同的霉心病判别模型中基于密度与特征光谱融合的模型总体判别率均高于分别基于密度、特征光谱的模型总体判别率,其中,基于密度与特征光谱融合的SVM模型总体判别率最高,为95.56%,而基于密度、特征光谱建立的SVM模型总体判别率分别为82.22%、91.11%,因此融合密度特征可进一步提高漫反射光谱对苹果霉心病的判别准确率,同时为开发基于漫反射检测原理的便携式苹果内部病害与品质一体化无损检测设备提供了可能。To solve the problem of low accuracy of near-infrared diffuse reflectance spectroscopy for moldy apple core discrimination,this paper proposes a multi-factor nondestructive detection method for moldy apple core by fusing density feature with diffuse reflectance spectroscopy.Based on the spectral acquisition platform to obtain the diffuse reflectance spectrum(200-1100 nm)information of 195 Fuji apples,the WLD-600 density meter was used to obtain apple density data.In addition,the standard normal variable transformation(SNV)was used to pre-process the spectral data.Moreover,the competitive adaptive reweighted sampling(CARS)and successive projection algorithm(SPA)were combined to extract the feature spectra related to moldy apple core.Four different models of moldy apple core discrimination were established by density,diffuse reflectance spectra,fusing density and diffuse reflectance spectra with partial least squares discriminant analysis(PLS-DA),Fisher’s discrimination,support vector machine(SVM)and least squares support vector machine(LS-SVM).The results showed that the overall discrimination rate of the model based on the fusion of density and feature spectra was higher than that of the model based on density and feature spectra in different moldy apple core discriminant models.The SVM model based on the fusion of density and feature spectra had the highest overall discrimination rate of 95.56%,while the overall discrimination rate of the SVM model based on density and feature spectra was 82.2%and 91.11%respectively.Hence,the fusion of density feature could further improve the discrimination accuracy of diffuse reflectance spectra for moldy apple core,and also provides the possibility of developing portable non-destructive testing equipment based on diffuse reflection detection principle for the integration of apple internal disease and quality.

关 键 词：苹果霉心病 特征融合 漫反射光谱 密度 无损检测 

分 类 号：S436.611.1[农业科学—农业昆虫与害虫防治]