Acquisition of mangrove spectrum properties and detecting the sensitive bands provide technology basis for inverse modeling and estimation by remote sensing for various indexes of mangrove. The typical mangroves of Guangxi Shankou Mangrove Reserve were taken for study objects, the standard spectrum curves of Bruguiera gymnorrhiza (Linn.) Savigny, Rhizophora stylosa, Kandelia candel, Avicennia marina, Aegiceras corniculatum, Spartina anglica and mudflat were gained by de-noising analysis of field-measured spectrum curves acquired by ASD FieldSpec 2. Analyzing the spectral characteristics and their differences, the authors found that the spectrum curves for various kinds of mangrove are coincident, the bands that appeared with reflection peaks and reflection valleys are basically identical, the within-class differentiated characteristics are comparatively small, the spectrum characteristics of mangroves are obviously different with Spartina anglica and mudflat. In order to gain the quantitative description for within-class differentiated characteristics of mangrove, space distance method, correlation coefficient method and spectral angle mapping method were used to calculate the within-class differentiated characteristics. The division accuracy of correlation coefficient method is higher than spectral angle mapping method which is higher than space distance method, and the result indicates that the spectrum differences of within-class mangrove and Spartina anglica are relatively small with correlation coefficients more than 0.995, and spectrum curve angle cosine values more than 0.95.