Research of the spectral characteristics of healthy and fusarium-infected wheat seeds variety Moskovskaya 56 by hyperspectral spectroscopy

Abstract

In this paper, a spectroscopic identification method is considered for determining the maxima and minima of spectral lines to identify pathogenic microflora in grain seeds. The paper presents the justification for the application of the method of hyperspectral imaging, in order to identify the disease fusarium in the seeds of soft winter wheat. Based on the graphs, it can see the general picture of the influence of the disease fusarium on wheat grains. There is a general decrease in the reflectivity of the grain surface. The strongest deviation of spectral lines is observed in the limit from 660 nm to 900 nm.

Investigation of Spectroscopic Peculiarities of Ergot-Infected Winter Wheat Grains

“…The estimated accuracy parameters of the models demonstrate the greater predictive accuracy for the PCA-SVM approach, adapted to solving nonlinear tasks, compared to PCA-LDA (see Table 3 ). Qualitatively similar results for predictive ability were noted in the work [ 31 ].…”

“…Such analysis made it possible to reduce data dimensions as soon as the highly impact spectral regions were identified. The analysis was performed within two stages [ 31 ]. At the first stage, a linear transition to the new orthogonal variables the so-called principal components (PCs) was carried out, and the scores were obtained for corresponding spectra.…”

“…The predictive power estimation for discrimination against infected grains (positive result) and healthy grains (negative result) using the above approaches was carried out based on such commonly used criteria as sensitivity, specificity, positive predictive value, negative predictive value and accuracy, which were defined as follows [ 31 , 34 , 35 , 36 ]: Sensitivity = T P /(T P + F N ), Specificity = T N /(T N + F P ), Positive predictive value = T P /(T P + F P ), Negative predictive value = T N /(T N + F N ), Accuracy = N corr /N tot . where T P is the number of true positive results, T N is the number of true negative results, F P is the number of false positive results, F N is the number of false negative results, N corr is the number of correctly classified samples, N tot is the total number of samples.…”

“…The predictive power estimation for discrimination against infected grains (positive result) and healthy grains (negative result) using the above approaches was carried out based on such commonly used criteria as sensitivity, specificity, positive predictive value, negative predictive value and accuracy, which were defined as follows [31,[34][35][36]:…”