Abstract: This study was carried out to detect nitrogen content in lettuce leaves rapidly and non-destructively using visible and near infrared (VIS-NIR) hyperspectral imaging technology. Principal Component Analysis (PCA) was performed on the average spectra to reduce the spectral dimensionality and the principal components (PCs) were extracted as the input vectors of prediction models. Partial Least Square Regression (PLSR), Back Propagation Artificial Neural Network (BP-ANN), Extreme Learning Machine (ELM), Support Vector Machine Regression (SVR) were, respectively applied to relate the nitrogen content to the corresponding PCs to build the prediction models of nitrogen content. R2p of the PLSR model for nitrogen content was 0.91 and RMSEP was 0.32. BP model of structure 5-2-1 with R2p of 0.92 and RMSEP of 0.21, ELM model of structure 5-10-1 with R2p of 0.95 and RMSEP of 0.19 and SVR model for nitrogen with R2p of 0.96 and RMSEP of 0.18, all got good prediction performance. Compared with the other three models, SVR model has the better performance for predicting nitrogen content in lettuce leaves. This work demonstrated that the hyperspectral imaging technique coupled with PCA-SVR exhibits a considerable promise for nondestructive detection of nitrogen content in lettuce leaves.