Real Time Sensor Image Fusion

"We propose a flexible, very-high performance sensor/data fusion and ground-map correlation processor for Enhanced Vision Systems. The approach utilizes adaptive-system architecture, with algorithms that are realized through an associative, self-organizingneural net implementation. The low cost emulation of this highly parallel network will utilize dynamically programmable, FPGA-based accelerator modules; this approach also removes the processor-memory bottleneck that typically limits high-speed processorperformance.The architecture includes preprocessing tailored to individual sensors on a local-area image basis. The geometric representation and registration operations retain the respective high-content data regions from disparate sensor types; an important exampleis in fusion of an infrared camera (azimuth-elevation data) with imaging mmw radar (range-azimuth data). Adaptive features, which range from threshold adjustments to statistical calculations, are realized through selective feedback loops from theprocessor output.The sensor/database correlation capability provides ground-map navigation update and hazard recognition. The fusion and correlation functionality provide the required flexibility for advanced avionics/EVS system integration. The ultimate result of suchevolution will be a dual use, multi-imager EVS suite, with optimally integrated database, and such added sensors as DGPS, EGPWS, ADS-B, and TCAS. The anticipated