Study on chlorophyll fluorescence spectrum in the application of the BP-ANN for diagnosing cucumber diseases and insect pests

The diagnosis model of the cucumber diseases and insect pests was established by laser-induced chlorophyll fluorescence (LICF) spectroscopy technology combined with Back Propagation Artificial Neural Networks (BP-ANN) algorithm in this research. This model would be used to realize the fast and exact diagnosis of the cucumber diseases and insect pests. Four kinds of in-vivo cucumber leaves (health, downy mildew, aphid, downy mildew and aphid) were used in this experiment, and detected the physiological information and chlorophyll fluorescence spectrum by the corresponding experimental apparatus and chlorophyll fluorescence spectrum collection system. The two hundred and nineteen samples were randomly separated into the calibration set and the validation set. Principal component analysis (PCA) was a method which had been widely used in the spectroscopic analysis for reducing the dimensionality of the noise reduction spectrum, and according to the model accuracy seven principal components (PCs) were selected to replace the complex spectral data. The noise of original spectrum was reduced by five methods, which include Savitzky-Golay smoothing (SG), First-order Differential Treatment (FDT), Fast Fourier Transform (FFT), Standard Normal Variate (SNV) and Wavelet Analysis (WA). According to the best diagnosis accuracy of calibration and prediction set, optimization results with every pretreatment method in different spectrum band were compared. Results showed that BP-ANN with input by the first seven principal components of full-band spectrum had the best identification capabilities and accuracy was 100% after the original spectrum noise was reduced by FDT. This research indicated that the method of BP-ANN had a good identification effect and could realize rapid diagnosis of the cucumber diseases and insect pests as a new method.