THREE-DIMENSIONAL HUMAN FACE RECOGNITION METHOD BASED ON MULTI-SCALE COVARIANCE DESCRIPTOR AND LOCAL SENSITIVE RIEMANN KERNEL SPARSE CLASSIFICATION

摘要

Disclosed is a three-dimensional human face recognition method based on a multi-scale covariance descriptor and a local sensitive Riemann kernel sparse classification. The method comprises the following steps: respectively carrying out automatic preprocessing on original G library set human face models and P test set human face models; according to the library set human face models and the test set human face models after same have been subjected to the automatic preprocessing in step (1), establishing a scale space and detecting multi-scale key points and extracting neighborhoods thereof; extracting d × d-dimension local covariance descriptors from a neighborhood of each key point neighborhood under each scale, and carrying out multi-scale fusion on the local covariance descriptors so as to construct a multi-scale covariance descriptor; and mapping the local covariance descriptors to a renewable Hilbert space, and proposing a local sensitive Riemann kernel sparse representation to classify and recognize a three-dimensional human face. By means of the present invention, the expression capability of a single-scale local covariance descriptor can be effectively improved, and at the same time, the locality of a multi-scale descriptor can be effectively used in a local sensitive Riemann kernel sparse classification.