Human face confirming method based on similarity measurement

Abstract

The invention discloses a human face confirming method based on similarity measurement. The method comprises the following steps that features of N pairs of human face images are extracted, each pair of human face features are subjected to PCA transform, and a feature vector of human face sample pairs is obtained; within-class weighed subspace is obtained through a weighted covariance matrix and subjected to weighted subspace projection, and feature vectors of the human face pairs obtained after projection are obtained; in combination of the knowledge of priori measurement, a priori similarity matrix and a priori distance matrix are obtained; a lagrangian multiplier method and a paired method are used for solving a target function of similarity measurement learning, and a measurement matrix is obtained; the similarity of human face pairs is calculated in combination with a similarity model, and therefore an optimal threshold value is obtained and used for human face confirmation. The problems that under the constraint-free condition, due to interference of expressions, postures, illumination, backgrounds and other factors, the difference of different human face images of the same person is large, and extracted human face feature vectors are greatly different are solved, and therefore the human face confirmation accuracy is improved.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a face confirmation method based on similarity measure. Background technique [0002] In recent years, face recognition technology under unconstrained conditions has been widely used in video surveillance, public security, e-commerce and other biometric identification, intelligent video analysis and computer vision systems. However, face images under unconstrained conditions are interfered by complex factors such as expressions, postures, lighting, and backgrounds, resulting in large differences in different face images of the same person. How to overcome the large variety of face images caused by the above factors It is one of the important problems that need to be solved urgently in the field of face recognition. [0003] Similarity measurement refers to using a given face sample pair to calculate the distance between the face pair feature vectors to judge the identity of the f...

AI Technical Summary

Problems solved by technology

[0005] However, due to the influence of complex factors such as facial expression, posture, illumination and background under unconstrained conditions, the intra-class variation of human faces is large, and the current mainstream face recognition technology is still unable to guarantee robustness. Accurate face recognition under the premise of security

Images