Olinda valencia orange leaves dry powder-like were taken as sample, and chemical analysis combined with technology of visible near-infrared spectroscopy (Vis/NIRS) was used, through the treatment process of second derivative spectrum of samples of the original spectrum and denoising (Noise). Meanwhile, method of partial least squares (PLS) and cross-validation were used to establish maths model of Zn concentration which applying band combination composited by 400-500 and 1201-1300 nm of characteristic wavelength band. The coefficient of establishing models is 0.9975, while the coefficient of correlation coefficient of prediction is 0.9920. The root mean square error of prediction (RMSEP) of cross-validation is 0.5868. Therefore, the means using visible near-infrared spectroscopy (Vis/NIRS) and the methods of cross-validation and PLS to establish the spectral correction model reflecting the Zn content in leaves and characteristic wavelength bands can detect the Zn content in citrus leaves quantitatively and quickly.