The unreliability of IU algorithms has severely limited more widespread adoption of IU systems in many domains. Techniques and approaches that promise significant advances in overall system reliability are of great interest. The reliability barrier is not likely to be broken through incremental improvements to known algorithms. Instead, research should seek methods to build reliable IU systems from unreliable components, or should pose different IU tasks that are both DoD relevant and can be solved reliably.In the future, image exploitation will be performed not by searching large tracts of terrain for significant forces and activities, but by monitoring and tracking the movements of all military equipment within a large, but limited, geographic area. This vision will be enabled by maintaining a vast repository of information about potential battlegrounds, such that image interpretation can be conducted in comparison to prior knowledge. Vast quantities of imagery will be collected routinely, and used to update the repository over as wide an area as is technically feasible. Image exploitation, then, could be performed with the aid of a wealth of data that is not available to today's IA. For example, in the near-term, the IU-assisted IA could prepare to track ground movement of important vehicles in a restricted geographical area (say a 5 km x 5 km battlefield, an airfield, a garrison, a bivouac site, or a road junction), by extracting salient information from imagery of that area that is routinely collected for whatever purpose. This task can be compared in some ways to air traffic control surveillance except it is focused on ground vehicles. As hostilities near, the IE system will use imagery from UAVs, which are operating in a flight pattern that provides frequent coverage (every hour or every minute). Near-real time intelligence reports are issued by the IA with the assistance of highly reliable IU tools.