The research on perilla identification algorithm based on FDA and SVM

To enable rapid identification of Perilla varieties and prevent the issue of counterfeit products being passed off as genuine, this paper selects three-dimensional fluorescence data from two Perilla varieties and proposes a feature selection method based on FDA and SVM for fluorescence data. This method successfully distinguishes among four Perilla varieties. First, the Delaunay triangulation interpolation technique is applied to eliminate the effects of Rayleigh and Raman scattering, thereby reducing their negative impact on the data. Next, the Savitsky-Golay (SG) convolution smoothing method is used to smooth the data and minimize noise interference. Following this, preliminary screening of the three-dimensional fluorescence data is performed, and FDA is employed for discriminant analysis of wavelet packet features, resulting in new variables through information fusion. Finally, BP neural networks and Support Vector Machines (SVM) are utilized to analyze these feature variables. The experimental results show that the neural network-based model achieves a training accuracy of 96.8% and a testing accuracy of 94.3%, while the SVM-based model reaches a training accuracy of 97.5% and a testing accuracy of 97.8%. These results demonstrate that the proposed method effectively identifies Perilla varieties and provides robust technical support for future research.