Specific Human Detection from Surveillance Video Based on Color Invariant Moments

This paper describes a detection framework to find a specific person from different videos by combining movement information with color invariant features. Special color invariant moment features based on color clustering are presented. Similarity measure method based on the color invariant moment descriptors is given. This method is applied in a practical application system and the experimental result demonstrates that the proposed features have a stronger ability of color based image region similarity discriminating than the traditional color features.