SBIR-STTR Award

Triton Systems proposes the utilization of novel computational techniques of sensor fusion and data visualization for enhanced situational awareness and anti-terrorism systems. The principal Investigator has successfully demonstrated this approach in demanding electronics manufacturing environments previously, making time-to-market should be relatively short. Since the manufacturing environments and the surveillance and reconnaissance data being processed in this program are both multidimensional time-stream data, we expect the same technique to be applicable to data sets from, for example UAVs. In Phase I, we will demonstrate the technology for sensor fusion and visualization of UAV or other reconnaissance data. This is a necessary first step toward large-scale implementation across many surveillance and intelligence gathering missions. Phase II will focus on improving the algorithms developed in Phase I, and validating the system with extensive data testing. Working with our partners, the Phase II will culminate in delivery of a complete set of application software applicable to many domains of sensor fusion and visualization associated with surveillance and reconnaissance. Phase III will culminate in delivery of a complete, and adaptable, system for field use. Successful completion of the Phase II will demonstrate the potential of this technique for wide-ranging applicability in all military and intelligence gathering operations.

The Learning Machine for Anti-Terrorism Applications utilizes a Neural Network architecture to implement a vector-matrix multiplication technique for data pattern recognition. This capability can be used to find potential data of interest within large data sets. A second Learning Machine can be used in conjunction with the results from the first to predict the next event based from a time-window of past events and bound that prediction with a probability (for time dependent data).

Keywords:
NEURAL NETWORKS, LEARNING MACHINES, VECTOR QUANTIZATION, SUPPORT VECTORS, COSINE METRICS, PERCEPTRON, POLYNOMIAL NETWORK