【摘要】 作物病毒病发生的早期检测能够为病害预警和综合防控提供关键信息。为探究病毒病早期无损检测的可行方法，本文分别使用高光谱成像技术和叶绿素荧光成像技术，以辣椒轻斑驳病毒侵染的辣椒幼苗、甘蔗花叶病毒侵染的玉米幼苗以及相应的健康幼苗为测试样品，对病毒侵染早期的辣椒和玉米幼苗开展了快速、无损检测。对高光谱数据及叶绿素荧光数据进行不同预处理后，使用随机森林(RF)、正交偏最小二乘判别分析(OPLS-DA)及支持向量机(SVM)算法构建了分类模型。结果表明，叶绿素荧光RF模型最适于辣椒显症前或显症后叶片带毒的判断，其训练集和测试集的准确率分别均为78.3%和80.0%;玉米显症前叶片高光谱数据经多元散射校正(MSC)预处理后的OPLS-DA模型最适于其带毒情况的判断，其训练集和测试集的准确率分别为90.0%和100.0%;叶绿素荧光OPLS-DA模型最适于玉米显症后叶片带毒的判断，其训练集和测试集的准确率分别为95.0%和100.0%。本研究为生产上使用高光谱或叶绿素荧光成像技术在早期检测作物幼苗病毒病提供了参考。更多还原

【Abstract】 Early detection of viral diseases in crops is critical for effective warning and integrated management. In this study, hyperspectral imaging(HSI) and chlorophyll fluorescence imaging(ChlFI) were employed to achieve rapid and non-destructive detection of early-stage viral infections in seedlings of pepper and maize. Pepper seedlings infected with pepper mild mottle virus(PMMoV) and maize seedlings infected with sugarcane mosaic virus(SCMV), along with their corresponding healthy controls, were used as experimental materials. Following various preprocessing techniques, classification models were developed using random forest(RF), orthogonal partial least squares discriminant analysis(OPLS-DA), and support vector machine(SVM). The results showed that the RF model based on ChlFI data was the most suitable for detecting viral infection in pre-symptomatic or post-symptomatic pepper leaves, with training and testing accuracies of 78.3% and 80.0%, respectively. For pre-symptomatic maize leaves, the OPLS-DA model using HSI data preprocessed via multiplicative scatter correction(MSC) achieved the best performance, with training and testing accuracies of 90.0% and 100.0%, respectively. For symptomatic maize leaves, the OPLS-DA model based on ChlFI data yielded training and testing accuracies of 95.0% and 100.0%, respectively. This study demonstrates the potential of HSI and ChlFI technologies for early, non-invasive diagnosis of viral diseases in crop seedlings, offering valuable reference for precision disease management in agricultural production.更多还原