In the present study, a 2-year maize field experiment with different fertilizer dose was conducted. The spectral response sensitive area of maize leaves in different levels was discussed using hyperspectral technology at booting stage. Based on the correlation analysis of original reflectivity and its first derivation with maize leaf nitrogen contents, prediction models were constructed. The results indicated that under different fertilizer dose spectral response sensitive areas of maize leaves were in visible band around 550 and 761~1 300 nm; under different levels which was in visible band around 550 nm; maize leaf nitrogen contents were significantly correlated with spectral reflectance and its first derivative in 470~760 nm band. Through further comparison and selection, the index prediction models built with spectral indices DSI(564, 681) and DSI(681, 707) were the best prediction models, the prediction accuracy were 93.43% and 93.39%, therefore nitrogen content of maize leaves could be effectively estimated.