Incremental Knowledge Assimilator Navy SBIR FY2009.1 Sol No.: Navy SBIR FY2009.1 Topic No.: N091-066 Topic Title: Incremental Knowledge Assimilator Proposal No.: N091-066-1221 Firm: UtopiaCompression, Corporation 11150 W. Olympic Blvd. Suite 680 Los Angeles, California 90064-1817 Contact: Joseph Yadegar Phone: (310) 473-1500 Web Site: www.utopiacompression.com Abstract: Mine countermeasures is an important aspect of Navy reconnaissance interests and reliable mine detection still remains a challenge. For automatic mine detection in underwater sonar applications, research has indicated that the identifying the context of the scene will improve the recognition performance. Further, there is a need to constantly update and modify the representative knowledge base with information from new target and background samples. Based on advances in machine learning and artificial intelligence and their sophisticated applications to intelligent imaging solutions, we propose an Incremental Knowledge Assimilator that can identify the environmental context in the scene and optimally incorporates new and relevant data samples into the knowledge base (classifier) in an incremental fashion with minimal memory and computation requirements. The incremental learning algorithm is designed to facilitate concept drift. Since the goal of the system is to learn perpetually, it lays higher stress on learning the patterns the each mode of the distribution (or each class) rather than focusing merely on the separation of classes. We present preliminary performance results to validate our methodology. Benefits: UC is determined to demonstrate success with the Navy as its initial customer and then expand its product offering to other service branches and agencies. Following phase II development, UC will work closely with both the Navy, and major vendors to migrate the trial system to initial deployment. UC will leverage its ongoing relationships with contractors such as Raytheon, with whom UC is collaborating on several SBIR projects (including a recently begun Phase II effort for the Navy related to visual knowledge discovery), and Boeing, to develop an effective and successful product transition plan for the proposed technology. UC's proposed technology will greatly benefit the Navy by enabling efficient and accurate mine countermeasures. Specifically, UC will 1)Automatically identify the environmental context in the scene using statistical texture models, 2) Develop an accurate and efficient knowledge representation system to recognize mine (possibly with multiple classes of targets) and fine tune knowledge representation system for each possible context in the scene to improve classification accuracy, 3) The system perpetually assimilates new and relevant information into the existing knowledge database in an incremental fashion, 4) The system also allows the knowledge representation to drift not only on the basis of most recent data but also on the basis of accuracy of IKA with respect to that data. Commercial Applications: The commercial potential for the proposed technology is notable. Not only does the Navy have a current need for the technology, shared by other military and intelligence divisions within the government, the technology proposed by UC also provides critical decision support capabilities needed by agencies such as DHS, NASA and others to meet their mission objectives. There is also a substantial need for UC's proposed technology in the broader commercial market. UC's knowledge mining technology will be of great use to the security and surveillance market in accurately and cost-effectively deploying systems for border, port and infrastructure monitoring. UC is very optimistic about the market opportunities in this area, given that the US security & surveillance market overall is projected to grow to more than $48B by 2009 [F04]. The Perimeter and Border Protection sector of the security & surveillance market alone is projected to generate over $8.5 billion in revenue over the next five years [FR06]. Another promising application area for UC's knowledge mining technology is the modestly expanding data mining market and related information systems. The data mining market is slated to expand by approximately 10% annually over the next few years according to Meta Group, Inc [B05]. With some modest re-factoring, UC believes that the generic classification algorithms can be retrained to address a variety of data types subject to mining. Some applications of potential interest include text analysis and mining, and bioinformatics. Return