看守所人脸识别（Prison house face recognition）.doc

看守所人脸识别（Prison house face recognition） Commonly used methods of face recognition (L) a method based on geometric features Geometric features were first used in face recognition. The face is made up of eyes, nose, mouth, chin, and so on, It is precisely because of the shape, size, and structural differences of these components that each individual face in the world is poor, Therefore, the geometric description of the shape and structure of these components can be used as an important feature of face recognition. The people The face is represented by a geometric feature vector, and the classifier is designed with the idea of hierarchical clustering in pattern recognition to identify the same This requires that the selected geometric feature vector has a certain uniqueness, and can reflect the differences between different faces, At the same time, it has a certain elasticity, so as to eliminate the influence of time span and illumination. The geometric feature vector is based on human face A feature vector based on the shape and geometry of an official, whose component usually includes the Euclidean distance between two points specified by the face Departure, curvature, angle, etc. (2) matching method based on elastic model The idea of the elastic model matching method is that each feature point on the human face is used as the reference point to form an elastic map The benchmark stores a representative characteristic vector, and uses hierarchical elastic graph to remove some redundant nodes, A sparse face description structure is formed. Recognition is performed by matching the test samples with the feature samples. elastic There are many methods for model matching, for example, a set of rectangular grid nodes is placed on the face image, and the characteristics of each node are presented Using the multi-scale Gabor amplitude feature at this node, the connection between nodes is represented by geometric distance, Thus, a face representation based on two-dimensional to