SBIR-STTR Award

This Small Business Innovation Research Phase I project explores the feasibility to develop a Supply Chain Optimization and Product Explorer (SCOPE) that includes a unique 3D Shape Search technology, and enables tighter integration across the extended enterprise and seamless flow of information up the supply chain to decision makers at the point when product parameters are being specified. The research and development objectives are to 1) Develop a 3D Shape Search Technology that will initially be geared towards search and retrieval of parts and products from a manufacturing supply chain. 2) Develop a SCOPE architecture to index 3D part and product information across a supply chain network. The architecture will be conceptualized and a representative model will be prototyped to show system as a whole, which would include the utilization of the 3D Shape search research and demonstration of search results. The 3D shape search algorithm will be developed by using the methodology of representing a 3D B-Rep part as a voxel model, which is an approximation of the 3D geometry by a set of cubic volume elements. The 3D shape search technology research will include the development of 1) Shape Representation using Landmark and Skeletal graphs. Skeletal graph representation method captures detailed level shape representation and hierarchical skeletal graph structure allows for local shape matching. 2) Shape comparison using low information loss, histogram representation of the graph to allow for quick and accurate graph comparison. 3) Intelligent Query Interface that will have a quick and novel way of representing the topology of the shape in the query, selection of feature of interest in the query part, and easily navigable cluster based search results interface. The proposed research, will allow application of the 3D shape search technology to bridge the gap realized by the industry, for a technology that will allow quick and accurate location of part and product data in real-time across the supply chain, thus enabling a cohesive extended enterprise and an optimized supply-chain. US manufacturers are currently faced with a critical gap in capability in current Supply Chain Management systems related to real-time search and retrieval of information from the supply chain at the point when product parameters are being specified leading to billions of dollars worth of obsolete parts inventory, locked up inventory on part information and wastage of valuable man-hours in non productive search and redesign work. SCOPE, along with its 3D search technology aims to address this technology gap, by enabling reliable and accurate access to part data stored in disparate formats across a supply chain, information that was hitherto inaccessible using traditional search technologies. Given a growing trend amongst companies in aerospace, manufacturing, medical equipments and other industries towards 3Ddata definition, the 3Dsearch technology provides a quantum leap in terms of accuracy, speed and relevance in the search and retrieval of information existing within the supply chain. Importantly the 3D shape search with its radical improvement in search and retrieval capabilities has the potential to create completely new markets and products, by becoming a foundation technology for advanced search systems having applications as diverse as identifying molecules for pharmaceutical or bio-technology research to face recognition for security agencies. SCOPE along with its unique 3D shape search technology will enable US companies in aerospace, manufacturing and medical equipments to save billions of dollars through better integration and optimization of their supply chain

This Small Business Innovation Research Phase II project will achieve higher retrieval accuracy for shape-based search for both the web and the enterprise. The proposed work in Phase II is to achieve higher retrieval accuracy supported by three key components: 1) pose determination for 3D models: bridging the space gap between 2D and 3D shapes by finding three intuitive and robust orthogonal orientations for 3D models; 2) 2D orthogonal view generation: representing a three orthogonal views along the pose orientations; 3) similarity measurement between 2D shapes: finding 2D and 3D shapes based on the user's query. A framework will be developed by focusing on three important modules: 1) 2D constraint detection and use of implied constraints with initial application in 2D and 3D views; (2) Enhanced multiple level-of-details in 3D representations, and (3) Human assisted system classification of large datasets.

Traditional options of finding part suppliers using catalogs, trade shows and prior business relationship limit the choice of suppliers. Current text-based search to find suppliers face challenges, such as context and language sensitivity, and is inadequate in overcoming the technological challenges posed by variations in how product or part information is specified across a global supply chain. The current effort proposes to use shape, which is the lowest common denominator, to link the OEMs and suppliers. This technology can also aid the current trend among companies in aerospace, automotive, medical equipments and other industries towards 3Ddata standards for fast retrieval, as it can provide a significant leap in terms of accuracy, speed and relevance in the search and retrieval of information. If successful, this technology can contribute significantly to research in areas where shape is important, such as bio-technology and pharmaceutical sectors, where rapid identification of molecules and their docking features help reduce time and cost involved in drug development. For the medical industry due to increased usage of CT scans and 3D imaging technologies, 3D shape search can be used for local feature identification in colonoscopy or other exploratory procedures, brain angiography, reconstruction, projection of malformation or location of polyps and ensure better and rapid diagnosis of disease. Development of methods for automatically parsing human sketches and determining constraints will enable many other research activities and broadly help in a more natural human machine interaction