Visible/near-infrared (Vis/NIR) spectroscopy was investigated for the fast discrimination of rice leaves with different genes and the determination of chlorophyll content. Least squares-support vector machines (LS-SVM) was employed to discriminate transgenic rice leaves from non-transgenic ones. The classification accuracy of calibration samples reached to 100%. Successive projections algorithm (SPA) was proposed to select effective wavelengths. SPA-LS-SVM discrimination model was performed, and the result indicated that an 87.27% recognition ratio was achieved using only 0.3% of total variables. The optimal performance of each quantification model was achieved after orthogonal signal correction (OSA). Performances treated by SPA were better than that of full-spectrum PLS, which indicated that SPA is a powerful way for effective wavelength selection. The best performance of quantification was obtained by SPA-LS-SVM model; with correlation coefficient (R) and root mean square error of prediction (RMSEP) being 0.902 2 and 1.312 1, respectively. Excellent classification and prediction precision were achieved. The overall results indicated that the new proposed SPA-LS-SVM is a powerful method for varieties recognition and SPAD prediction. This study supplied a new and alternative approach to the further application of Vis/NIR spectroscopy in on-field classification and monitoring.