Dealing with the Nonlinear Battlefield

THE collection and analysis of overhead imagery has come a long way over the past 50 years or so. During the Cold War, analysts scrutinized overhead imagery at light tables, pouring over static scenes with which they were familiar and searching for significant changes. Analysts were typically assigned to count missile silos, look for large-scale troop movements, and spot long-range bomber deployments.

Since the Yugoslav wars in the early 1990s, the West has had few large-scale troop movements to monitor. We have attempted, sometimes successfully, to use imagery to find conventional military units—Serbian tanks and artillery, for example—but we now also recognize the importance of being able to track the movements of small groups of people and individual vehicles. In doing so, we could potentially locate safe houses and arms caches and provide general situational awareness in an environment where the enemy would disperse its forces to avoid U.S. airpower and where no clear boundary separated small enemy units from civilians and friendly forces. The battlefield has become “nonlinear.”

The Yugoslav conflicts saw for the first time the U.S. beginning to use unmanned aerial vehicles (UAVs) as a significant tool. UAVs collected video, mostly of poor quality, over broad areas and kept objects in view until a weapon could be brought to bear. The deployment of UAVs implied the potential availability in future conflicts for huge amounts of dynamic imagery and the need for new analytic methods to process these data.

Our concern over the nonlinear battlefield and the need to deal with very small enemy units have been driven home in the current conflicts in Iraq and Afghanistan. The desire to detect the emplacement of an improvised explosive device or the transport of materials to and from a bomb factory has led to the concept of persistent surveillance of a village or city, including tracking the movements of vehicular and foot traffic and determining the location of enemy facilities. A further goal of persistent surveillance is to fuse collected data with other intelligence collection capabilities to determine the "human terrain" of social interactions and hierarchies.

Persistent surveillance implies continuous, full-motion, high-definition video over a broad area. However, transmitting such highly detailed visual data requires very large bandwidth, which cannot be streamed through the downlink from a UAV. Much of the current research in compression techniques is driven by the explosion of the Internet, but the information content that typical Internet users seek is not fully consonant with the needs of the military.

As described in the article From Video to Knowledge, a Livermore system called Persistics permits real-time analysis of visual data by compressing unimportant (unchanging) features by a thousandfold and highlighting features of interest. Persistics capitalizes on the Laboratory’s deep knowledge of sensor technology, our capabilities in high-performance computing, and our understanding of the Department of Defense's needs. Unlike other compression techniques that degrade quality, delete critical military data, or fail to achieve the needed compression, Persistics preserves relevant information in the scene, namely pedestrian and vehicular motion. Analysts assisted by real-time automated exploitation algorithms can obtain good situational awareness without having to examine every image manually. It is a highly creative solution to an important military problem and has been embraced by the Department of Defense.

A further requirement in dealing with the nonlinear battlefield is providing timely intelligence at the tactical level. After the Yugoslav conflicts, U.S. military strategists saw the advantage of transmitting just-acquired visual data directly to squad and platoon leaders instead of analyzed intelligence flowing down from the highest levels to brigade and company commanders. This once-controversial concept, now widely accepted, forces the issue of providing the sergeant or corporal with data that are easy to interpret. These soldiers consistently ask for full-motion video. Again, we see a mismatch between the need for data, the lack of associated bandwidth, and the number of analysts required to draw timely conclusions. Addressing this issue is a Department of Defense priority and the focus of one of the Laboratory’s mission enhancement initiatives. As exemplified by the Persistics team, the creativity associated with Livermore’s multidisciplinary culture, and its capabilities and infrastructure, will have an enormous effect on the American soldier.