An Automated Visual Knowledge Discovering (VKD) System

As imagery sensing is among the essential capabilities in the future combat systems, the anticipated dynamics of the future battlespace will require a greatly increased level of automation in analysis and processing of imagery data. Most of the image analysis tasks including indexing and retrieval need or would benefit from a full explanation of the scene, consisting of the locations and scales of detected objects and their relationships to one another. For many DoD data, however, there are challenges to achieving this capability, including the large variability in scene configuration and object attributes, and the difficulty in incorporating geometric constraints between objects. UtopiaCompression Corp. (UC) proposes a novel Visual Knowledge Discovery (VKD) technology that can affectively address these issues. The system performs automated scene analysis, inference, and annotation using a novel hierarchical and contextual image model. The model provides a full explanation of the scene including the scene hierarchy and the relationship between objects, and is capable of representing a variety of scenarios with large variability in scene structure and number of objects. The proposed technology will significantly enhance the capabilities of the DoD in annotation, indexing, and knowledge mining of large image databases.