Human Identification Based On Multi-biometrics Under Surveillance Video

With the rapid development of science and technology in the modern society, social security and information security shows the unprecedented importance. Human identification as a guarantee of the security has become a key social problem and hot research area of great concern. With its advantages of the high security, stability and effectiveness,biometric identification technology has become one of the most active research direction of identification field. But with the development of research,we found that single mode biometric identification has showed some inherent defects in such aspects as identification accuracy, universality and security. In order to solve this problem, researchers put forward identification scheme using multi-biometrics data fusion, has received widespread attention. Recently multi-biometric recognition technology is still in the exploring study stage and is not yet mature, needs further research and perfection.This paper design a identification fusion scheme based on multiple biometric fusion of human face and gait aiming at the key technical problems of biometrics. Take real-time requirements and the identification accuracy into consideration and puts forward fast and efficient biometric feature extraction algorithm and matching algorithm, adaptive fusion of face and gait based on decision level is proposed, improves the accuracy of identification.In the aspect of face recognition, Adaboost cascade classifier method is used for face detection, by using different haar- like features to detect face of different angles, extracting face feature with nonnegative matrix decomposition, using Euclidean distance matrix for face recognition and classification.As in gait recognition, this paper put forward Dynamic Energy Matrix DEM(Dynamic Energy Matrix) based on the gait energy image, and classify each test object according to the similarity function through the principal component analysis. Compared with the traditional gait energy image, on the one hand, dynamic energy matrix can reduce the shooting conditions change profile caused by transverse deformation during construction, on the other hand, it highlights the gait uniqueness of different objects. So it is a better description of gait characteristics of objects. In aspect of multiple biometric fusion, this paper proposes a adaptive fusion recognition method of multi-angle face and gait based on decision level, respectively extract face and gait characteristics and gain the Euclidean distance of test of image to the training image array collection, using nonlinear programming to each biological feature distance matching value for adaptive allocation to get the optimal joint matching points, to achieve the best information fusion results. Because the fusion weights changes with different test objects and angles, thus it has certain real-time performance and adaptability. The experimental results show that the proposed adaptive fusion features carry the most discriminating power compared to any individual biometric and other static fusion rules like MAX and SUM.