Abstract: The techniques of correlation analysis and principal component analysis are utilized to study leature microstructure parameters extracted from SEM photos of the three kinds of soil. The study shows that message superposition is clear and severe among microstructure parameters for feature analysis, and that is same for field analysis. Therefore, it is reasonable to construct a lower parameter matrix to substitute the original higher parameter matrix for microstructure. The minority representative principal components can approximately synthesize the overall meaning of microstructure according to the desired value of contribution by the method of principal component analysis, and this makes extensive use of mierostructure theory in soil mechanics expediently and potentially.