Subspace-based incremental learning face recognition method

Abstract

The invention discloses a subspace-based incremental learning face recognition method, which comprises the following steps: 1, pre-processing a face picture; 2, generating more than one group of training sample subspace and one group of test picture subspace; 3, comparing the similarity degree of the test picture subspace and the more than one group of training sample subspace; and 4, incremental learning: and if the similarity degree of the test picture which is identified as a j*th class and the sample picture which represents the face of the j*th class is more than a set threshold value, regulating the group of sample subspace with a base change method to generate one group of new sample subspace. The subspace-based incremental learning face recognition method has the advantages that the incremental learning is realized, contacted information can be fully utilized in the identification process, and an identification result is improved.

Description

technical field [0001] The invention relates to the technical field of pattern recognition, in particular to a subspace-based incremental learning face recognition method. Background technique [0002] In the past two decades, face recognition technology, as an efficient biometric identification technology, has been increasingly valued by academia and industry. Compared with other technologies, face recognition technology has the advantages of naturalness, high acceptability, and not easy to be detected by faces during recognition, so it has great application potential in video surveillance, criminal investigation, entertainment, military and other fields. [0003] Early researchers often performed face recognition by selecting and extracting geometric features of face images. However, because the extraction of geometric features is very difficult, and it is very sensitive to changes in expression, lighting, and shooting angles, later researchers tend to use feature extract...

AI Technical Summary

Problems solved by technology

Its disadvantage is that multiple corresponding subspaces must be generated separately for each face picture, which means that even two photos of a person with little difference must be represented by two sets of subspaces. Even though the subspaces of the two groups are themselves quite similar

Since the time spent on recognition and comparison is proportional to the number of sample subspaces, when the number of acquired samples increases a lot during the incremental learning process, the time spent on recognition will increase accordingly.

Images