基于高光谱成像技术的南果梨酸度无损检测方法  

Non-destructive Testing Method for Acidity of Nanguo Pear Based on Hyperspectral Imaging Technology

作　　者：张芳[1] 邓照龙 田有文[1] 高鑫 王开田 徐正玉[2] ZHANG Fang;DENG Zhaolong;TIAN Youwen;GAO Xin;WANG Kaitian;XU Zhengyu(College of Information and Electrical Engineering,Shenyang Agricultural University,Shenyang 110161,China;Hefei Treasury,Hefei 230031,China)

机构地区：[1]沈阳农业大学信息与电气工程学院,沈阳110161 [2]合肥市财政局,合肥230031

出　　处：《沈阳农业大学学报》2024年第2期231-239,共9页Journal of Shenyang Agricultural University

基　　金：辽宁省教育厅基础研究项目(JYTMS20231285)。

摘　　要：南果梨是一种重要的水果品种,其酸度是评估果品质量的重要指标之一。然而,传统的南果梨酸度检测方法通常需要破坏性采样和化学分析,不仅耗时费力,而且容易导致样品污染和浪费。因此,旨在探索一种基于高光谱成像技术的无损检测方法,以实现对南果梨酸度的快速、准确、无损检测。首先,采集室温20℃下不同贮藏天数南果梨的高光谱数据,其光谱波长范围为400~1000 nm,并且通过理化实验测量南果梨样本的可滴定酸;其次,采用多元散射校正(multipli⁃cative scatter correction,MSC)、标准正态变换(standard normal variate,SNV)、Savitzky-Golay平滑滤波等多种方法对光谱数据进行预处理,建立偏最小二乘回归模型(partial least squares regression,PLSR),选择出建模效果最佳的预处理方法,结果显示MSC方法效果最优;然后结合连续投影算法(successie projection algorithm,SPA)提取特征波段,在700~900 nm范围内确定9个特征光谱变量;最后,以提取出的9个特征光谱变量作为输入矢量,分别建立PLSR模型、极限学习机(extreme learning machine,ELM)模型以及遗传算法(genetic algorithm,GA)和粒子群算法(particle swarm op⁃timization,PSO)优化的BP神经网络模型。研究结果表明,基于MSC预处理和SPA算法特征提取的PSO-BP模型预测精度最高,效果最好,预测集决定系数R^(2)_(p)=0.911,RMSEP=0.032。可见,基于高光谱成像技术的SPA-PSO-BP模型可用于南果梨酸度的检测,为南果梨的品质评价提供参考。Nanguo pear is an important fruit variety,and its acidity is one of the important indicators for evaluating fruit quality.However,traditional methods for detecting acidity in Nanguo pear often require destructive sampling and chemical analysis,which is not only time-consuming and laborious,but also prone to sample contamination and waste.Therefore,a non-destructive testing method based on hyperspectral imaging technology was explored to achieve rapid,accurate,and non-destructive detection of acidity in Nanguo pear.Firstly,the hyperspectral data of Nanguo pear stored for different days at room temperature of 20℃was collected,the wavelength range is 400-1000 nm,and the titratable acid of Nanguo pear samples was measured through physical and chemical experiments;secondly,multiple methods such as multiple scatter correction(MSC),standard normal variation(SNV),Savitzky Golay smoothing filtering were used to preprocess spectral data.A partial least squares regression(PLSR)model was established,and the best preprocessing method was selected.The results showed that the MSC method had the best performance;then,combined with the continuous projection algorithm(SPA),feature bands are extracted,and 9 feature spectral variables are determined in the range of 700-900 nm;finally,using the extracted 9 feature spectral variables as input vectors,a PLSR model,an extreme learning machine(ELM)model,and a BP neural network model optimized by genetic algorithm(GA)and particle swarm optimization(PSO)were established respectively.The research results indicate that the PSO-BP model based on MSC preprocessing and SPA algorithm feature extraction has the highest prediction accuracy and the best performance,with a prediction set determination coefficient R^(2)_(p)=0.911 and RMSEP=0.032.It can be seen that the SPA-PSO-BP model based on hyperspectral imaging technology can be used for the detection of acidity in Nanguo pear,providing reference for the quality evaluation of Nanguo pear.

关 键 词：高光谱成像技术 南果梨 酸度 BP神经网络 PSO-BP模型 

分 类 号：TS255.3[轻工技术与工程—农产品加工及贮藏工程]