No abstract

Human matching between different fields of view is a difficult problem in intelligent video surveillance; whereas fusing multiple features has become a strong tool to solve it. In order to guide the fusion scheme, it is necessary to evaluate the matching performance of these features. In this paper, four typical features are chosen for the evaluation. They are the Color Histogram, UV Chromaticity, Major Color Spectrum Histogram, and Scale-Invariant Features (SIFT). Quantities of video data are collected to test their general accuracy, robustness, and real-time applicability. The robustness is measured under the conditions of illumination changes, Gaussian and salt noises, foreground errors, resolution changes, and camera angle differences. The experimental results show that the four features bear distinctive performances under the different conditions, which will provide important references for the feature fusion methods.

Through analyzing the low resolution video captured by a single camera fixed on the air condition, this paper proposes an approach that can automatically estimate the person's location and recognize the person's identification in real time. Human location can be obtained by smart geometry calculation with the knowledge of the camera intrinsic parameters and living experience. Human recognition has been found to be very difficult in reality, especially when the person is walking at a distance in the complexity indoor conditions. For optimal performance, we use the shape feature gait energy image (GEI) as the basis, since it isn't sensitive the noise. Then we extract more efficient features using the histograms of oriented gradients (HOG) and do the dimensionality reduction by the coupled subspaces analysis and discriminant analysis with tensor representation (CSA+DATER), Finally the classical Bayesian Theory is used for fusion of the result of HOG and the result of CSA+DATER. The proposed approach is tested on our lab database to evaluate the performance of the human location and recognition. To verify the robust of our human recognition approach especially, CMU MoBo gait database is used. Experimental results show that the proposed approach has a high accuracy rate in both human identification recognition and location estimation.

No abstract