Identification of Mildew in Brassica rapa var.chinensis Seeds Based on Hyperspectral Imaging Technology
In order to distinguish healthy and mildewed Brassica rapa var.chinensis seeds,the spectra of healthy and mildewed Brassica rapa var.chinensis seeds were obtained by hyperspectral imaging technology,and the discriminant model was established.Support vector machine discriminant(SVM-DA)model was established based on original spectra and 9 kinds of pre-treated spectra.It was found that the model based on Firstderivative pre-treated spectra performed best,and the accuracy of modeling set and prediction set were 95.87％and 95.74％,respectively.In order to remove redundant or unnecessary information,genetic algorithm(GA)was used to screen the bands of the original spectra and various pre-processed spectra,and SVM-DA model was established again.Among these models,FD-GA-SVM-DA model had the best performance,and the accuracy of modeling set and prediction set reached 97.71％and 96.81％,respectively.In conclusion,it was feasible to identify healthy and mildewed Brassica rapa var.chinensis seeds based on hyperspectral technique.
万方数据
维普
