Deep-learning-based regression model and hyperspectral imaging for rapid detection of nitrogen concentration in oilseed rape (Brassica napus L.) leaf

Deep-learning-based regression model composed of stacked auto-encoders (SAE) and fully-connected neural network (FNN) was used for the detection and quantification of nitrogen (N) concentration in oilseed rape leaf. SAE was applied to extract deep spectral features from visible and near-infrared (380-1030 nm) hyperspectral image of oilseed rape leaf, and then these features were used as input data for FNN to predict N concentration. The SAE-FNN model achieved reasonable performance with R2P = 0.903, RMSEP =0 .307% and RPDP = 3.238 for N concentration. Results confirmed the possibility of rapid and nondestructive detecting N concentration in oilseed rape leaf by the combination of hyperspectral imaging technique and deep learning method.