Object Retrieval Scheme Using Color Features in Surveillance System

In this paper, we have described the object retrieval scheme based on color for video surveillance that is influenced by the different light changes and overlapping/non-overlapping view cameras setting. The proposed video retrieval scheme separates object into top and bottom, and extracted dominant colors from each region. Each dominant color includes hue, saturation, value in HSV space and proportion of hue color. In addition, it uses the various threshold values and pre-defined weights based on the experiment and processes the similarity measurement to order the search results. Therefore, our retrieval scheme provides the delicateness and the robustness in varying surveillance environmental conditions. As well, it can be applied in real-time surveillance system.