基于可见/近红外光谱的水蜜桃糖度无损检测方法优化研究  

Optimization of nondestructive testing method for soluble solid content of peach based on visible/near infrared spectroscopy

作　　者：张小斌[1] 朱怡航 赵懿滢 陈妙金 孙奇男 谢宝良 冯绍然 顾清[1] ZHANG Xiaobin;ZHU Yihang;ZHAO Yiying;CHEN Miaojin;SUN Qinan;XIE Baoliang;FENG Shaoran;GU Qing(Institute of Digital Agriculture,Zhejiang Academy of Agricultural Sciences,Hangzhou 310021,China;Fenghua Peach Research Institute,Ningbo 315502,Zhejiang,China;Beijing Sunshine Yishida Technology Co.,Ltd.,Beijing 100020,China)

机构地区：[1]浙江省农业科学院数字农业研究所,浙江杭州310021 [2]宁波市奉化区水蜜桃研究所,浙江宁波315502 [3]北京阳光亿事达科技有限公司,北京100020

出　　处：《浙江农业学报》2023年第7期1617-1625,共9页Acta Agriculturae Zhejiangensis

基　　金：浙江省重点研发计划项目(2021C02052)。

摘　　要：糖度是水蜜桃最重要的品质指标之一。水蜜桃糖度无损检测方法主要借助可见/近红外光技术来实现,但该技术受到的制约因素较多。本文综合探究分析各类条件下近红外光谱技术在水蜜桃糖度无损检测中的应用,分析比较不同品种组合建模与测量方式,以提高水蜜桃糖度无损检测的精确性。研究以浙江宁波奉化本地的3个品种水蜜桃样本为研究对象,获取水蜜桃光谱数据进行建模分析。首先利用可自行建模的H-100型无损糖度检测仪进行实验,分析结果表明,不同品种构建的近红外光谱模型的均方根误差(RMSE)和决定系数(R^(2))不同,其中新玉模型的结果最佳,RMSE为0.22,R^(2)为0.98;各类品种建立的模型对各自品种的数据分析结果最佳,而对其他品种数据的分析结果较差;数据种类越多,数据量越大,模型越优秀,3种品种混合模型的R^(2)高达0.92。其次,采集奉化本地3个品种的水蜜桃样品,使用久保田K-SS300、ATAGO PAL-HIKARi 10及H100(自建模型)3款无损设备分别进行同部位检测,并将检测值与ATAGO PAL-1的有损检测值进行相关性分析。结果显示,对于每个品种,H100设备均具有最高的准确性。采集奉化本地的混合品种水蜜桃进行不同深度的糖度检测,结果表明,H100设备获取的糖度值能更好地反映水蜜桃整体的糖度水平,而久保田和ATAGO只能反映水蜜桃外部区域的糖度水平,无法表征水蜜桃整体的糖度水平。最后,本文还探究了两个水蜜桃品种在不同硬度条件下对H100设备检测结果的影响,结果表明,水蜜桃硬度的下降会较大地影响水蜜桃糖度的无损检测。因此,利用H100型无损检测仪结合有效的模型能较好地规避近红外光谱技术受到制约因素的影响,同时为无损检测水蜜桃糖度提供一定的参考。Soluble solid content(SSC)is one of the most critical quality indexes of peach.Nondestructive detection of SSC in peaches is mainly realized by visible/near-infrared(VIS/NIR)spectroscopy,but it has some limitations.In this paper,the applications of NIR spectroscopy in the nondestructive detection of peaches under different circumstances were comprehensively explored and analyzed.The modeling and detection methods of different cultivars were analyzed and compared to improve the accuracy of nondestructive detection of SSC in peaches.In this study,three local cultivars of peaches collected from Fenghua were selected as the research object,and the spectral data were acquired for modeling.Firstly,the self-modeling H-100 nondestructive sugar content detector was used to conduct the experiment.The results showed that the root mean square error(RMSE)and R^(2) values of the NIR models established on different cultivars were different,among which the Xinyu model obtained the best results,with RMSE of 0.22 and R^(2) of 0.98.The model based on single cultivars had the best predictive ability to detect the respective cultivar,but it performed worse when detecting other cultivars.The more cultivars of peaches,the larger amount of data,the better the model.The R^(2) of the model based on three cultivars was as high as 0.92.Secondly,Kubota K-SS300,ATAGO PAL-HIKARi 10,and self-modeling H100 were used to detect the same part of three local cultivars of Fenghua peaches,respectively.The correlation analysis was conducted between the values predicted by the nondestructive detectors and those obtained by the destructive ATAGO PAL-1.The results showed that the H100 detector had the highest predictive accuracy for all the cultivars.Next,SSC detection at different depths of mixed cultivars of Fenghua peaches.The results showed that the SSC levels obtained by the H100 device could better reflect the overall quality.In contrast,Kubota K-SS300 and ATAGO PAL-HIKARi 10 could only reflect the SSC levels of the external areas.Finally,the influe

关 键 词：水蜜桃 无损检测 糖度 模型 近红外 

分 类 号：S126[农业科学—农业基础科学]