SBIR-STTR Award

This Small Business Innovation Research (SBIR) Phase I project describes a powerful new approach to GIS data-mining for large image databases. The current inefficiencies in extracting information from these databases are primarily caused by the limited ability to extract baseline geographic information (such as road networks) from images. The proposed system addresses this need by combining feature extraction efficiencies with the ability to refine queries efficiently. The feature extraction system reduces the labor required for extracting information from imagery and the query refinement system incorporates user-driven and data-driven strategies to focus on interesting and relevant discoveries. The proposed system will also provide solutions for multi-media applications containing imagery, web-based applications, medical image repositories, and other applications. Geo-spatial information technology is rapidly becoming a cornerstone of scientific research, environmental modeling, local government planning, and federal government security programs. The proposed system also has commercial potential for other application domains such as medical imaging

This Small Business Innovation Research Program Phase II project will develop a highly innovative data mining software tool that is capable of mining imagery and spatial information stored in a database management system (DBMS). Billions of dollars have been spent in converting the world's vast supply of paper maps into digital, geographically referenced, data for geographic information systems (GIS) applications because location matters in almost every instance of decision-making for both government agencies and private sector businesses. The proliferation of relational, spatial, and now visual data from high-resolution satellites, all stored in a common DBMS architecture, offers organizations the opportunity for knowledge discovery in databases; however, the technical challenges of maintaining, navigating, and mining these data are formidable. Current workflow approaches are disjointed and exclusive of image data. The product resulting from this project will allow all of the data to be queried and mined in a holistic workflow approach yielding potential useful discoveries through its primary innovations not presently available in data mining software; 1) Seamless integration of data mining and feature extraction workflows, 2) Mining content of high-resolution earth imagery stored in spatial databases, 3) Cleanup of GIS databases, and 4) Advanced query generation and data mining technology.