Face binary feature extraction method based on sparse constraint

The invention discloses a face binary feature extraction method based on sparse constraint. On the basis of a classic binary feature learning algorithm, a projection matrix with a row sparse structure is induced to be generated on the basis of sparse constraint terms, and in the projection process, the contribution degree of main features in original features can be improved, and the influence of secondary and redundant features can be reduced. In an actual scene, secondary features which are easy to be interfered by noise are often secondary features, and main features are not easy to be influenced by the secondary features. Therefore, the face features extracted by the method have better robustness and stability, and the method has practical application value for face recognition. 本发明公开了一种基于稀疏约束的人脸二值特征提取方法。在经典的二值特征学习算法的基础上，本发明基于稀疏约束项，诱导产生具有行稀疏结构的投影矩阵，在投影过程中，能够提高原始特征中主要特征的贡献度，减少次要和冗余特征的影响。实际场景中，容易被噪声干扰的往往是次要特征，主要特征不易受其影响。因此本发明提取到的人脸特征具有更好的鲁棒性和稳定性，对人脸识别具有实际应用价值。