Application of Blind Sources Separation in plant leaves classification

This paper discussed the application of Blind Sources Separation (BSS) in plant leaves classification. Firstly, collection of two different types of plant leaves was performed using the Nexus-870 Fourier transform infrared spectroscopy, and wavelet analysis was adopted to compress the immense sample data, thus accelerating the data processing speed. Then the BSS algorithm FastICA algorithm was used on the compressed spectral data to increase the difference between the different signals. Finally, BP neural network algorithm was used to achieve the classification of plant species. Experiments showed that processing data in near-infrared spectroscopy through BSS can not only improve the speed and accuracy of BP neural network, but also enhance its classification correctness, and the classification results with the proposed method was satisfactory.