The invention discloses a data enhancement method for robust face identification. According to the method, a face shielding processing method is included, if an input face image has no shielding, a standard formed by all the no-shield samples is selected as a shielding mask, if the face image corresponds to one type shielding, a corresponding shielding mask is selected for reconstruction; noise is added in a manual mode, and noise enhancement processing is carried out; fuzzy set image enhancement is carried out; HsV equalization homomorphic filtering is carried out, through image spectrum analysis, influence of illumination and reflectivity on a pixel gray value is processed, the image is reduced to be a more clear one, and more details are realized; image generation is accomplished based on triangular face multi-attitude characteristics. The method is advantaged in that influence of shielding, noise, fuzziness, illumination, expression and attitude factors can be prevented in a face identification system, a problem of insufficient training data in the face identification system can be solved, and model generalization capability is improved.