TUAV "video" data can support numerous missions - tactical unmanned aerial vehicles

Military Intelligence Professional Bulletin, Jan-March, 2003 by John Dugan, Donald Wurzel

The most common missions described for tactical unmanned aerial vehicles (TUAVs) are in support of situational awareness and targeting. The principal sensor developed to provide data to satisfy these needs on low- to medium-altitude-endurance unmanned aircraft is the video camera. Video cameras can provide very useful movies of areas of particular interest that are crucial for observing the surroundings and for manually detecting and tracking targets of high interest. The newsprint is replete with stories of their utility in support if intelligence, surveillance, and reconnaissance (ISR) missions.

Example UAV systems and products being developed by the U.S. Army have been described by Colonel Knarr, et al, in a recent article in the ISR Journal. (1) In almost all tactical-level systems developed and fielded to date, the video data are reviewed manually by a human operator, and targets of interest are detected unaided by computer processing. Such processing, when coupled with an enhanced onboard navigation capability, could dramatically increase the accuracy of UAV-derived data for improved precision targeting support and provide additional battlefield preparation data as well.

The navigation subsystems on present and near-future small UAVs are designed primarily as an aid to safety of flight. They do not provide sufficient navigational accuracy to exploit available digital imagery. That is, the data are not very useful for mensuration of the imagery onto a geodetic grid, which is essential for determining Global Positioning System (GPS) coordinates for precision-guided munitions (PGMs). Good examples are the current Air Force Predator and the Army HunterUAVs which do not provide accurate, digital metadata for mensurating the imagery. They provide a good picture, but GPS coordinates cannot be derived from the video. The navigation shortfalls inherent in the Predator and Hunter include inaccurate measurements of both the camera location and attitude during the collection of any particular image frame. If these measurements and the range to the target were known accurately, the geo-position of targets in the imagery could be calculated precisely. The best accuracy that can be achieved on operational systems today is hundreds of meters in three dimensions. Such poor accuracy does not enable the new family of small PGMs that all the Services are developing or producing, nor does it enable other important mission products discussed herein.

There are a number of possible techniques that could be employed to establish the accuracy needed for better precision targeting. One could use a high-accuracy inertial navigation system (INS) and a laser-range finder, wherein the target location is calculated directly from measurements of the camera location and attitude and the range to the target. This approach is called direct geo-referencing. It is expensive for a tactical UAV, presents laser eye-safe issues, and has not yet yielded desired accuracy. Another current alternative would be to calculate the geo-location of a manually detected target by positioning it relative to ground control points (GCPs) which are recognizable features in the image, and then find these GCPs in a reference image that is geodetically correct.

This approach currently is being developed for Joint Service Imagery Processing System (JSIPS) workstations. Unfortunately, the relationship between the video-derived image and the reference image is not always one-to-one because of changes that may have occurred between the times that the two were collected (for example, winter versus summer, new construction, battle damage) or changes in the relative viewing geometries of the source platforms. In addition, there may well be circumstances where suitable reference images simply are unavailable.

Finally, this is not a rapid process, so it does not address the time-critical strike issue. There are other techniques that have emerged recently that have been successfully demonstrated on imagery from surrogate UAVs, although they have not yet transitioned beyond the Science & Technology (6.2/6.3) stage of development. These approaches key on inexpensive commercial-off-the-shelf (COTS) technologies and a specific mix of hardware and software wherein good geodetic positioning of the camera and advanced processing algorithms are used to calculate the camera attitude and the relative target position. Thus, this approach can be considered a variant of direct geo-referencing.

This approach uses a rather inexpensive INS that provides good camera positioning via augmented GPS techniques, but relatively inaccurate attitude solutions due to the use of inexpensive inertial-measurement units (IMUs). The higher noise levels in the COTS IMU can be compensated by one of several techniques being developed under Office of Naval Research (ONR) support. These techniques recalculate the onboard estimates of camera-attitude angles by one of several different algorithms. One technique even takes advantage of features in the images, and it is perhaps of higher risk because it is so unique, but it potentially provides significantly higher payoff because the accuracy of the results, in principle, are independent of the stand-off range. This technique has been applied to imaging data collected during fleet exercises using the Airborne Remote Optical Spotlight System (AROSS), also developed under ONR support. Absolute positioning accuracies of better than 5 meters from 3-kilometer range have been demo nstrated, which are adequate for targeting of today's and near-future weapons. Which one of these many approaches will be the best for tactical systems has not yet been determined, and this topic of geodetic referencing of the imagery remains a fundamental one for resolution by further research and development.