Research and Implementation of Identity Recognition Based on Multi-Feature Fusion

Face recognition has been applied in numerous identification systems, however the constraint on the distance of faces and video sensors suppresses its development. The practical facts turn out that the performances in terms of accuracy and latency decline along with the distance increase. This paper focuses on identification at distance and proposes a multi-feature fusion identity recognition algorithm (MFIR) based on the Principal Component Analysis. Multi-feature combines height, clothing and functional face features from contour extraction for target identity at a distance. This combination makes the weighted eigenvalue a better representation of the characteristics for the identified object. We conduct experiments on the proposed algorithm with real data. The results show that the accuracy of MFIR compared with face-feature identification outperforms in long-distance scenario.