基于光谱技术的大豆豆荚炭疽病早期鉴别方法

为更好地指导农户进行植物病害防治,提高大豆豆荚的商品性,减少损失,需要运用快速有效的方法来进行大豆豆荚炭疽病的早期检测。该文应用可见-近红外光谱技术结合连续投影算法(SPA)和最小二乘支持向量机(LS-SVM),实现了大豆豆荚炭疽病的早期快速无损检测。对194个大豆豆荚样本进行光谱扫描,通过不同预处理方法比较,建立了大豆豆荚炭疽病早期无损鉴别的最优偏最小二乘法(PLS)模型。同时应用主成分分析(PCA)和连续投影算法(SPA)分别了提取最佳主成分和有效波长,并将其作为LS-SVM的输入变量,建立了PCA-LS-SVM和SPA-LS-SVM模型,以样本鉴别的准确率作为模型评价指标。试验结果显示PCA-LS-SVM和SPA-LS-SVM模型都获得了比较满意的准确率,且SPA-LS-SVM模型的准确率最高为95.45%。研究表明,SPA能够有效地进行波长选择,进而使LS-SVM模型获得较高的鉴别率,说明应用可见-近红外光谱技术鉴别大豆豆荚炭疽病是可行的。这为进一步应用光谱技术进行大豆生长对逆境胁迫的反应提供了新的方法,为实现大豆病害的田间实时在线检测提供参考。

Early detection of soybean pod anthracnose based on spectrum technology

In order to predict soybean pods anthracnose early and effectively, Visible/near infrared (Vis/NIR) spectra technology combined with successive projections algorithm (SPA) and least square support vector machines (LS-SVM) was investigated for the rapid and non-destructive discrimination of such soybean disease. Total 194 samples were collected, the best partial least squares (PLS) model was established comparing with the different pretreatment methods. The principal component analysis (PCA) was used to extract the best principal components (PCs), and the SPA was used to extract the effective wavelengths. The best PCs and the effective wavelengths were respectively used as input variables for the PCA-LS-SVM and SPA-LS-SVM disease detection models. The validation set indicated that both models had acceptable accuracy rate, especially SPA-LS-SVM model has an accuracy rate of 95.45% in predicting fungal infections. According to the results, SPA was a powerful way for the effective wavelengths selection, and Vis/NIR spectroscopy was feasible for the identification of colletotrichum truncatum on soybean pods. There is a potential to establish an online field application of early plant disease detection based on visible and near-infrared spectroscopy.