Photogrammetry for defense applications differs in many respects from traditional approaches, but continues to drive the discipline’s development.

The phrase “defense applications of photogrammetry” may spark visions of war heroes studying blurred aerial photographs through a magnifying glass or image analysts in a darkened, armored headquarters poring over sequences of satellite images to deduce enemy deployments and tactics. Appropriately, drawing conclusions from such imagery with the aid of photogrammetry illustrates the powerful capabilities of ingesting images from multiple sources and relating them mathematically to the Earth’s surface.

In fact, defense work has been a golden thread throughout the history of photogrammetry. As detailed in “GEOINT: A Historical Look at Photogrammetry and Remote Sensing Advancements,” page 10, such applications have influenced the discipline’s development for more than a century. This article examines some of the most recent breakthroughs in defense-related photogrammetry applications, highlighting some of the software advances that may find their way into commercial products in the near future.

Software Functionality
There’s a growing acceptance among military analysts that historical approaches to image analysis, whereby photographic prints were annotated and accompanied by written reports, must be superseded by the use of geographic information system (GIS) databases. To do so means graphics recorded on imagery can be stored in terms of their ground coordinates. Only then can graphics from one image be superimposed on a new image, making it feasible to compare images accurately through time. Moreover, imagery and related reports must be accessible to multiple image analysts and to those to whom they report their findings, thus groups of specialists can bring their expertise to bear.

As a result, sensor modeling is critical. Whether the sensor’s position and attitude are read from the metadata captured with the image—direct georeferencing via Global Positioning System (GPS)/Inertial Measurement Unit (IMU) technology—or the image is related to the ground through triangulation with ground control points or by identifying and measuring corresponding points in the image and an existing image database, the mathematical model relating image and ground positions is crucial.
 

Once imagery is on-screen, each image with its appropriate sensor model, multiple images and maps can be viewed simultaneously and related geometrically. Much of the image interpretation performed by an analyst is interactive—i.e., the process isn’t highly automated, and the tools required are primarily focused on simplifying tasks and improving ergonomics to reduce fatigue and mistakes during the long shifts required to cope with massive streams of incoming images. Image analysis software must cope with images that are often enormous and perform operations, such as roaming and rotation, with tremendous smoothness and speed. The operator expects to be driven around the image systematically, either in a standard pattern or along a user-prescribed path. The zoom or “magnifying glass” function is critical. Much of an analyst’s work involves image comparison, which requires the ability to flicker between images or look through a “porthole”—e.g., a color composite reveals the high-resolution panchromatic image beneath.

As an analyst works, he or she needs counting tools—e.g., to total the number of vehicles that can be identified—and extensive annotation tools to indicate points and delineate lines and areas, as well as to show particular features. Hyperspectral and multi-spectral tools are increasingly important. A simple tool used extensively is “find-in-scene,” whereby the spectral signature of an object of interest is used to identify similar pixels throughout an image or an area of interest within it—e.g., to delineate wet areas that may inhibit ground transportation, or for target acquisition. At all times the analyst must be able to record ground coordinates of points of interest to the maximum possible accuracy commensurate with the imagery and metadata or triangulation being used.

Thus, defense-related image analysis differs in many ways from commercial image analysis. In addition to the obvious difference—ground control is often hard to come by in conflict situations—there are many dichotomies between the sets of applications. Defense forces have access to a much wider range of sensors than their civilian counterparts. Projects tend to be smaller, limited to specific areas. Speed is of the essence, with results needed in hours rather than weeks or months. Accuracy is important in critical areas of the workflow— e.g., targeting—but in certain situations end users must be content with whatever results can be generated reasonably from the available imagery and metadata. Though the deliverables in individual projects may be used immediately then no longer required, national defense agencies are amassing gigantic databases and libraries of imagery, digital terrain models, buildings and finished products, including worldwide coverage in many cases. Defense personnel creating and using geographic information derived from photogrammetry are trained and motivated differently from their civilian counterparts. And the requirements of national security reduce the visibility of imagery and technical data to personnel passing background checks, in some cases limiting the pool of photogrammetric talent available for urgent research, development and production.

Sensor Modeling
For years, photogrammetry in defense applications has suffered from a large variety of sensor types, many of which weren’t fully intended for photogrammetric use. Intelligence and reconnaissance operations often underemphasized the requirements for geometric stability and sufficient field of view for mapping and higher accuracy applications. The situation has challenged photogrammetrists to model the wide variety of sensors and attempt to account for their characteristics for practical needs, such as using error propagation.

Today, new sensors are adopted more rapidly, even in the commercial market. Half a dozen or more new sensors are brought into use each year in the defense segment. Such sensors tend to be designed by various vendors and branches of government, and there’s little standardization of the resulting support data.

However, given the host of reference imagery available in global databases, a pressing requirement is software that enables the image analyst to pull reference imagery, maps or terrain models from the databases and automatically, accurately and rapidly match the new imagery to the reference and thereby orientate the incoming images. The imperative of availability means that images of different types sometimes have to be matched—e.g., electro-optical to radar—necessitating novel algorithms that encompass feature and area matching at various resolutions. The small field of view of some of the sensors causes a “soda straw” effect, whereby a user loses his or her reference point while viewing, for example, a video stream. To solve the problem, recent developments include real-time video mosaicking and georeferencing, whereby tie points between video frames are identified and measured.

In addition, the U.S. government is moving into light detection and ranging (LiDAR) technology, an area in which there has been intense commercial development in recent years. For example, the new BuckEye change-detection sensor integrates commercial-off-the-shelf LiDAR and digital camera components to meet particular requirements. Similarly, the use of radar imagery acquired at different times has become a powerful tool for detecting small change—e.g., the planting of mines and improvised explosive devices. And extensive efforts are under way to combine human intelligence—e.g., sketches of the interiors of buildings—with imagery and databases.

Ongoing Integration
For more than a century, a remarkable stream of research bears witness to the role of defense as a primary driver of photogrammetry development. Thus, it’s no surprise that defense photogrammetry differs from normal commercial or civilian government practices.

However, the worlds of traditional image analysts and defense photogrammetrists, who call themselves geospatial analysts, are merging. Indeed, such integration is reflected in ongoing software trends and continuing growth in the range of commercial and government imagery available. These resources exist alongside massive global databases of existing maps, terrain models and georeferenced imagery. Geospatial analysts enjoy rapid, automated triangulation, often matching new imagery to reference sources, incorporating sensor models of several types, and emphasizing multisensor approaches and rigorous error propagation.

Beyond these extensions of the traditional triangulation process, defense requirements are behind many tools and techniques, from new high-resolution satellites through unmanned aerial vehicle (UAV)-borne imagers to sensor fusion and a plethora of information technologies. These are exciting times for image analysts and geospatial analysts, with better solutions at their fingertips than ever. And with more than a century of well-documented history, the remote sensing industry has the hindsight to ensure these innovations will diffuse into the commercial photogrammetric world as soon as possible.

  See  Featured Images
 
  Subscribe to Earth Imaging Journal

  Go to Home Page