In recent years, as video monitoring system is becoming more digitalized and more intelligent, more and more attention has been paid on intelligent monitoring technologies. Because in specific circumstances, specific persons always wear uniforms, it is possible to distinguish the role of the objects by discerning the persons' dressing. Thus, dressing identification technologies are quite significant in the intelligent video monitoring system, in which detecting and tracing the specific objective based on specific roles could obtain better performance.This paper first analyzes and compares existing classical image classification algorithms, and by focusing on the properties of the intelligent surveillance system and the specific circumstances described here, adopts a support vector machine-based image classification algorithm to implement the analysis and comparison of human dresses, and experiments are conducted under realistic environments. The experiment results show the good performance and real-time property of the human dress recognition algorithm being used in role recognition of the intelligent surveillance system. |