As the quote above most eloquently illustrates, a significant problem facing the intelligence community at this time is the disparity between the vast quantities of geospatial and other intelligence data acquired by various sensors/methods and the capacity of traditional human analysis and data extraction techniques. This problem will only get worse as new sensors and systems come online (FIA, WorldView, SBR, etc.) to meet post 9-11 needs for more ubiquitous and persistent surveillance.

A wide range of new technologies are urgently needed to address this problem. This is best illustrated by the "intelligence pyramid" shown below.

At the base of the pyramid you have the overwhelming amount of source from multi-modal literal and non-literal (SAR, hyperspectral, etc.) collection systems from a variety of platforms (satellite, airborne, ground, etc.). The development of automated processing techniques including automatic feature extraction (AFE), automatic target recognition (ATR) and tracking, automatic scene description (ASD), change detection, etc. is the first step in dealing with the voluminous amount of sensor data. Next, robust strategies for fusing the collected information from multiple sensors as well as other intelligence tradecrafts (signals, human, etc.) are needed for integrated multi-intelligence (multi-INT) analyses. After this, smart database technologies including content-based information retrieval (CBIR), software intelligent agents, and knowledge or semantic information mining are needed to extract the most useful, relevant, and/or highest-priority information.

Automated processing techniques combined with content-based information mining systems can, optimistically, assist in processing between 90 to 95 percent of the raw data volume. The remaining 5 to 10 percent is still a huge volume that must be analyzed. High performance human-computer interfaces (HCI) are needed for manipulating, comparing, interrogating, and further processing of the most relevant data and information. Visualization tools are needed to focus analyst attention and assist in rapid visual scene information extraction and correlation with intelligence information from multiple tradecrafts. In addition, these tools should adaptively learn what information is most relevant for a particular analyst and this information needs to be dynamically fed into the lower-level CBIR systems to assist in information and knowledge retrieval.

The multi-level integrated technology-based approach described above will allow human analysts to rapidly identify relevant information and focus their attention on the most urgent items. In turn, this will facilitate knowledge discovery and the exploitation of geospatial intelligence rather than its extraction.

The Center for Geospatial Intelligence (CGI) is comprised of a unique multi-disciplinary team of researchers that, together, have extensive experience across all levels of the intelligence pyramid. The CGI is engaged in significant R&D activities at UMC in the areas of satellite and airborne remote sensing (multispectral, lidar, SAR, video surveillance), advanced image processing techniques (AFE, ATR, ASD, scene registration, scene understanding), large dataset visualization, urban scene reconstruction, intelligent database and information retrieval, and detection/characterization of underground facilities. Examples of past and current research in these areas are provided by the links on the left. By leveraging these multi-disciplinary research skills, the center conducts leading-edge research focused on geospatial intelligence needs critical for national security, homeland defense, and military combat support.