Waterlogging seriously affects the yield and quality of cotton，whereas the waterlogging tolerance as a complex agronomic trait is difficult for quantification on a large scale using traditional methods. Seeking an efficient method for identifying cotton genotypes showing waterlogging tolerance might guarantee cotton breeding for quality and yield. Hyperspectral technology is a nondestructive and efficient method for crop detection can quickly identify cotton seedlings showing waterlogging resistance. In this study，the plant height，stem diameter，leaf area，and total root length of 27 cotton varieties were measured under hypoxia stress or normal oxygen conditions. Portable spectrometer was used to obtain its spectral image，analyze the waterlogging characteristic curve and extract six spectral indexes related to pigment，moisture and nitrogen. Multivariate analysis methods such as principal component analysis，stepwise regression，systematic clustering，and correlation analysis were adopted to construct a cotton seedling waterlogging tolerance identification model based on hyperspectral images. Principal component analysis showed that 10 individual indicators of waterlogging tolerance might be simplified into four principal components，which were further simplified into the D value of the comprehensive evaluation coefficient of waterlogging tolerance by considering the contribution rate and weight of the principal component and the normalized feature vector. Stepwise regression analysis was performed to establish the optimal linear regression equation of D value and 10 single index waterlogging tolerance coefficients：D=0.161+0.220 SD+0.068 LA+0.358 PSSRa +0.404 I1+0.292 NDSI （R2=0.9849，SD：stem diameter，LA：leaf area，PSSRa：pigment specific simple ratio a，I1：infrareed index 1，NDSI：normalized diffiferent stress index）. Cluster analysis of 27 cotton genotypes suggested four grades of waterlogging tolerance，including extreme waterlogging tolerance，waterlogging tolerance，medium waterlogging tolerance，and waterlogging sensitivity. The correlation analysis showed that stem diameter was significantly positively correlated with pigment ratio index a， pigment ratio index b， plant nitrogen spectral index， normalized difference stress index， and infrared index 1，but had no significant correlation with normalized difference water index. The total root length was significantly negatively correlated with pigment ratio index a， plant nitrogen spectral index， and normalized difference stress index，but had no significant difference with normalized difference water index and infrared index 1. There were no significant correlations between leaf area and plant height and the other 9 individual indicators. Collectively，this study highlighted the hyperspectral imaging technology as a useful method to identify the waterlogging tolerance of cotton.