What are the recognition methods face recognition technology
2022-10-28 00:00:00
MIT laboratory Turk Turk and pant (Pentland) & put forward by the other;Characteristics of face & throughout;Method is undoubtedly during the same period, the most famous face recognition methods.Followed by a lot of face recognition technology is more or less relationship with the characteristics of face, now face has been associated with the association of Normalized quantity (Normalized Correlation) method, a performance test datum algorithm of face recognition.
the fisherman face recognition method
bell mill, such as hu jintao proposed Fisherface face recognition method is another important achievement of this period.This method firstly adopts principal component analysis (PCA) for image apparent feature dimension reduction.On this basis, adopt the method of linear discriminant analysis (LDA) transformation after the dimension reduction of principal components in order to obtain & other;As far as possible big divergence between classes and as far as possible little divergence within class & throughout;.This method at present is still one of the mainstream method of face recognition, produced a number of different varieties, such as zero space method and subspace discriminant model, enhance the discriminant model and direct LDA sentenced to don’t method and some recent improvements based on kernel learning strategy.
elastic graph matching method
the basic idea is to use an attribute graph to describe the face: attribute graph vertices represent key facial feature points, its properties for the corresponding feature points in the resolution and direction of local characteristics & ndash;—Gabor transform 12 characteristics, known as the Jet;The attributes of the edge is the geometric relationships between different feature points.For arbitrary input facial image, elastic graph matching through an optimal search strategy to locate the key to a predefined number of facial feature points, extracting their Jet characteristics at the same time, the attributes figure of the input image.At last, by calculating its similarity with known face attributes figure to complete don’t process knowledge.This method has the advantage of both retain the global structure of facial features, also key local characteristics of the human face modeling