Abstract: A total of 14 shape characteristic parameters including the perimeter area, the maximum diameter, Elliptic and regional from cotton50, Xinluzao26, Xinluzao33 were investigated based on image processing, central tendency and dispersion of multivariate statistical analysis. The parameters were then used as the index of species identification. With step discrimination analysis methods of the SPSS, the identification model was established for three varieties, and three linear classification function were obtained. Through verification, the detecting accuracy of the forecasting group, including cotton50, Xinluzao26 and Xinluzao33, were 90%, 80% and 100% respectively. The results show that the recognition rate of these three cotton seeds is high.