Biometric Identification Study In Intelligent Video Surveillance System

In recent years, the intelligent video surveillance system develops rapidly in the field of computer vision, and it has become an important research direction in the field. Biometric identification technology, taking advantage of the human body inherent characteristics, belongs to identity authentication computer technology, which is widely used in the field of banking identity authentication and judicial identification. Biometric identification technology in intelligent video surveillance system is the biometric identification technology is applied to intelligent video surveillance systems to construct an automatic identifying and processing system. The biometric identification in intelligent video surveillance system is more complicated than the static image-based biometric identification, due to the interference of light, environment noise and body movement. Thus, the analysis and research of interference factors in intelligent video surveillance system with biometrics has great value. Biometric identification methods in intelligent video surveillance system are research priorities in this thesis.The purpose of this thesis is to analyze the biometric identification algorithms in the intelligent video surveillance system to improve the accuracy of the body identification, which contributes to the real application of intelligent video surveillance system.In the surveillance video, biometric identification has a characteristic of non-contact and non-invasive. And the features used in identity authentication include the face characteristic, gait feature and body posture. In this thesis, these three characteristics are analyzed and the corresponding feature recognition algorithms are raised. The main contributions are summarized as follows:1. The face recognition technology has been studied. First of all, the processing methods of capturing face image in the intelligent video surveillance are analyzed and the recognition beneficial facial image is obtained by preprocessing. Then through analyzing and improving the traditional LBP algorithm, the improved LBP face recognition algorithm is proposed. By calculating the mean and variance within the face region, the four-value model of this region is obtained. Through experiments, the improved method has an improved recognition rate and good robustness properties comparing with the conventional LBP algorithm. Finally, the improved LBP operator is applied to the intelligent face recognition system.2. A new gait feature extraction method is proposed based on Kinect through the research on the gait feature. Three kind of gait feature can be extracted by the tool of Kinect, namely: the joint angles information of legs, the walking stride feature, and the three-dimensional body contour descriptors. The depth obtaining principle and the skeletal obtaining principle based on Kinect are described in this thesis, and the three-dimensional body contour is got by the coordinate transformation in Kinect. Finally, the nearest neighbor classifier and K-nearest neighbor classifier are used in the experiments, which show that the proposed gait recognition method based on Kinect is effective with the recognition rate of 84%.3. A tentative body gesture recognition method is put forward, and the characteristic of the body is analyzed which could be used for the theoretical basis and condition of identification. Human body features are defined and then the body posture characteristics are extracted by using the skeleton tracking function in Kinect. Finally the standard Euclidean distance classifier, reciprocal weighted Euclidean distance classifier and decision tree classification are respectively used in the experiments, which show that the human body gesture recognition method is effective with the highest recognition rate of 87%.