Face recognition is what recognition method

2022-07-26 00:00:00

First, the geometric feature of face recognition methods: geometric features can be the shape of the eye, nose, mouth, etc and the geometric relationships between them, such as the distance between each other.These algorithms to identify speed, need memory is small, but the recognition rate is low.

the second, based on the characteristics of the face (PCA) face recognition methods: characteristics of face is face recognition method based on KL transform, KL transform is a kind of optimal orthogonal transformation of image compression.High dimensional image space after KL transform to get a new set of orthogonal basis, keep one of the important orthogonal basis, from the base can be r. vishny low dimensional linear space.If the face in the lower dimensional linear space projection has divisibility, the projection can be used as the recognition of characteristic vector, this is the feature of face.These methods require more training samples, and is based on the statistic characteristics of image gray level.There are some characteristics of the modified method of face.

the third, face recognition method of neural network, the neural network input can be reduce the resolution of the face image and local autocorrelation function and partial texture of second order moment, etc.This kind of method is also need more training samples, and in many applications, the number of samples is very limited.

4, elastic graph matching face recognition methods: elastic graph matching method in two-dimensional space defines a for face deformation usually has certain distance invariability, and USES the attribute diagram to represent a face, topology of any vertex all contain a feature vector, is used to record information face near the vertex positions.Grayscale characteristics and geometrical factors were combined in the method, can allow when comparing images exist elastic deformation, in overcoming the expression changes affect recognition has received the good effect, at the same time for a single man is no longer need multiple samples for training.