Clutter Removal and Substantially Improved Submarine and Mine Detection Through Affordable Navy STTR FY2005 Sol No.: Navy STTR FY2005 Topic No.: N05-T025 Topic Title: Clutter Removal and Substantially Improved Submarine and Mine Detection Through Affordable Proposal No.: N054-025-0216 Firm: Brainlike Surveillance Research, Inc. 1081 Camino del Rio South, Suite 209 San Diego, California 92108 Contact: Robert Jannarone Phone: (619) 299-5139 Web Site: www.Brainlike.com Abstract: Magnetic detection of submarines, mines and other objects of interest has continued to improve in both sensitivity of the sensors and reduction in the complexity and cost of sensor devices. Sensors alone, regardless of how far science is able to improve their abilities, will not solve the problem. The Navy will have to find processing solutions capable of quickly and cost-effectively highlighting objects of interest from magnetic sensor information. Brainlike Surveillance Research, Inc, proposes to develop a novel estimation system that adapts sensor data for improved target identification -- automatically, efficiently, and adaptively. Output from the system shows anomalies clearly, removes background clutter effectively, adapts to changing conditions automatically, and improves results from complementary classifiers substantially. The proposed system will run a novel, efficient kernel algorithm that learns baseline parameters and correlations automatically and continuously. The kernel operates efficiently, to the point of being deployable on remote sensor arrays. Along with the proposed system, Brainlike offers novel analysis methods that focus on increasing target identification rates, reducing false alarms, and automating the analysis process, resulting in reduced costly incidents, false alarm costs, and staffing requirements. Benefits: This proposal outlines a simple, robust, and easy to implement magnetic sensor monitoring system that identifies threatening submarines and mines that significantly reduces Anti-Submarine Warfare and Mine Warfare monitoring costs and greatly increase the likelihood of target detection. Current state-of-the-art systems focus primarily on detecting known signatures of threatening targets, require many person hours of effort to create and maintain, and require significant system resources to operate. Brainlike proposes a solution which is uniquely positioned to address all targets (known and unknown) due to its ability to determine normal activity and highlight deviations from normal expected sensor behavior. Beyond its ability to detect known and novel targets over and above current solutions, Brainlike's system has proven its ability to be applied to a problem in less than half the time required to implement competing solutions. System resource requirements will be minimal due to the fact that Brainlike's analytics require very little processing power and can be distributed at end nodes. By distributing Brainlike analytics at the end nodes, data can be assessed at the source, substantially reducing bandwidth previously required to transmit insignificant information up stream to central processing locations. Given its lightweight, compact ability to detect novel and changing targets, the military can use the proposed solution across all of its systems and networks, providing a sizable return on investment. Return