Signal Processing to Expose Concealed Targets through Registration

We propose a comprehensive program to develop FOPEN registration algorithms, evaluate their performance, and implement the algorithms in a software package suitable for delivery to AFRL. An accurate and robust registration capability is a requirement formany FOPEN automatic target detection algorithms including coherent change detection, multi-look detection, and multi-sensor terrain delimitation/context integration. In Phase II, we will develop new registration algorithms and demonstrate theireffectiveness to improve the performance of existing FOPEN ATD/C algorithms. We will measure performance as a function of registration accuracy. We have designed our program to leverage and extend a registration architecture developed in ongoing programsat ALPHATECH and will couple this architecture with our understanding of UHF and VHF phenomenology. Within this architecture we propose to assemble a suite of algorithms that includes geocoding using Digital Elevation Maps, pixel matching (correlation andmutual information), feature matching, and the simultaneous registration of multiple images and creation/modification of a Digital Elevation Map. The image products that will be registered under this effort include VHF imagery, UHF imagery, X-Band imagery,and orthophotos from the Keystone '97 experiment and other imagery, e.g., FOPEN ATD imagery, provided by the government.