Eigen-Similarity Integral (ESI) - A New Concept for Invariant Image Similarity Detection

Most of today's precision guided weapons use Global Positioning System (GPS) signals to gain improved accuracy. But, as Operation Iraqi Freedom recently showed, foreign militaries have equipment to locally jam GPS signals. Another effective method ofautonomous navigation is necessary to ensure mission success. Similar to rudimentary terrain recognition used in early cruise missiles, the matching of terrain video images with onboard digitized terrain information can be used to accurately provide aplatform with effective positional awareness. The team of 21st Century Systems, Inc. and the University of Nebraska at Omaha is pleased to address the first step in achieving this worthy goal. Our proposed research focuses on developing a mathematicallysound approach for image similarity detection and extraction. The project is aimed at the development of a methodology that is suitable for various military weapon system applications such as target identification and autonomous navigation of unmannedvehicles. The method is based on the Eigen-similarity of a set of image features. We call this concept the Eigen-Similarity Integral (ESI). The advantages of an ESI-based method for comparing and matching the dominant components of images includecomputation effectiveness, uniqueness, flexibility, and conciseness. The ability to perform a computationally effective comparison of real-time images with images or video from a library is a key enabling technology for many military and commercialactivities. Militarily speaking, this capability could be used for automated target recognition, automated target tracking, and autonomous navigation for unmanned weapon systems. Certainly, a navigation or targeting system based solely on GPS is neitherrobust nor fault-tolerant. An ESI-based navigation system would provide that advantage. An ESI-based software application would accelerate the exploitation of intelligence and UAV camera imagery and ground moving target indicator (GMTI) video. Using theESI-based image discerning capability of GMTI video, contextual map data, and intelligent agents, this application of our research could assist with target recognition, target tracking, and detection through clutter (i.e., trees). Many non-militaryapplications involving video monitoring would benefit strongly from the core concept: facility and physical security, traffic monitoring, and others. Indeed, the Homeland Defense arena could make excellent use of the results of this research.