A Single Training Sample Face Recognition Method

Abstract

The invention discloses a face recognition method for a single training sample. The method comprises the steps of: 1), inputting a human face sub characteristic training sample material; 2), structuring a plurality of training samples; 3), extracting P sub characteristics of each training sample; 4), appointing an optional training sample, calculating the difference value of two images in the training sample according to the P sub characteristics by using a measuring module, structuring a P-dimensional characteristic data vector v of the sample, when two photos in the training sample represent a same person, the response value of v is r which is equal to 1, otherwise r is equal to 0; 5), obtaining a training result data set of machine learning according to the step 4); and 6), inputting two human face photos to be recognized and compared, and then recognizing. According to the method, the recognition capability of face sub characteristics is achieved by structuring a multi-training-sample set of the face sub characteristic in advance; and the face recognition method for the single training sample is achieved by using a sub characteristic recognition fusion technique.

Description

technical field [0001] The invention relates to a single training sample face recognition method, which belongs to the technical field of face recognition technology (FaceRecognition Technique, FRT). Background technique [0002] Face recognition technology is the most representative and challenging important technical direction in the field of biometrics. Face recognition refers to the recognition of one or more faces from static or dynamic scenes based on known face sample databases using image processing and / or pattern recognition techniques. [0003] Face recognition will face two situations, one is that the training samples are sufficient, and the other is the situation where the training samples are not very sufficient. Some face recognition methods are difficult to obtain ideal recognition results when the samples are relatively scarce. For example, in applications such as ID card verification and passport verification, each person has only one face image for recogni...