Human recognition based on head-shoulder moment feature

Aimed at the shortcomings of the traditional visual surveillance system, the automatic detection and recognition algorithm of human are studied in intelligent monitoring system. This paper uses the moment eigenvector of head-shoulder´s contour as the back-propagation (BP) neural network´s input for human identification by building the 2D model of human head-shoulder. Because of adopting the partial contour shape of the human rather than whole features, it has a better classification on solving the issue of the loss of property arising from human occluded easily in practical applications. At the same time, the mapping relation of ´feature- class´ is established by error back-propagation neural network classifier, which completes identification of human. The experiment´s result shows that this method is effective, and it has strong robustness.