By Rick Jones, general manager at Sanborn (www.sanborn.com), Colorado Springs, Colo.

Recent introductions of digital aerial cameras and sensors offer geographic information system (GIS) users a greater number of choices for populating geospatial databases and feature layers with remotely sensed data. Aerial imagery users now can choose between digital or film imagery. For those choosing a digital approach, a long list of acquisition and processing options must be considered to choose the right image for the application at hand.

Digital Benefits and Specifications
Before the digital aerial camera era, data capture meant flying over an area with a film camera, developing the film, scanning it into a digital format, and processing the raw digital data to extract features or create GIS layers. With digital cameras, data often go directly from an airborne hard drive to a processing stream on the ground, an approach that can cut processing time by weeks. Additionally, digitally acquired images are first generation, so there is no loss of data quality—a potential problem during film scanning.

Digital imaging also saves time and money by enabling the collection of multiple data sets—often panchromatic (black and white), multispectral color and near-infrared—in a single pass, something that is rarely possible with a film camera.
Once the decision has been made to use a digital aerial acquisition technology, it’s important to consider several characteristics of the imagery that will be collected before commissioning the project. GIS users familiar with digital satellite imagery will find the specifications of digital aerial systems are nearly identical to those of satellites.
Spatial resolution: What level of detail is required? One-meter resolution is common from aerial and satellite systems, but a spatial resolution of a few inches can be achieved with digital airborne cameras. In the raw data, the spatial resolution and pixel size are the same

.
Spectral resolution: The newest digital aerial cameras can acquire panchromatic, or one-band imagery, as well as multispectral imagery. Multispectral imagery typically is acquired as separate bands in the green, blue, red and near-infrared portions of the electromagnetic spectrum. Some commercial satellites, like Landsat and SPOT, offer additional bands that go further into the middle-infrared and thermal portions of the electromagnetic spectrum. In addition, some satellite and airborne hyperspectral imaging systems break the electromagnetic spectrum into dozens or multiple dozens of bands of smaller wavelengths.

Accuracy: As in film acquisition, accuracy refers to how close a pixel or individual feature in the image is to its actual location on the ground. GIS users should match the accuracy of imagery with their GIS data to ensure everything lines up well. If the imagery is more accurate than the GIS layer, users may need to rectify the GIS data.
 

Temporal resolution: This refers to how often another image can be acquired over the same spot on the ground. For satellites, the revisit time is dictated by orbit. Landsat, for instance, has a fixed revisit cycle of about 16 days. Newer satellites have typical revisit rates between three and five days. Conversely, a plane can fly over the same area every day, or even more frequently, providing greater temporal resolution. Frequent revisit rates generally are preferred to monitor rapidly changing situations, such as floods or fires.

Radiometric resolution: This refers to the sensor’s sensitivity to variations in reflectance. Radiometric resolution is usually specified by the “bits” of data that can be collected, usually either 8 or 12 in most digital aerial sensors. Sensors that collect more bits are able to classify and render images that have subtle variations in reflectance, such as those with deep shadows. Highly reflective, or bright surfaces such as sandy deserts or snow-covered ground, also can be resolved in detail. High radiometric resolution is preferable for GIS applications because it directly affects the ability to extract GIS layers and features from dark and bright ground surfaces.

Processing the Data
The key to extracting useful GIS information from digital image data lies in the processing. Many manual and automated techniques are available to exploit digital imagery. There are two ways to extract GIS features from imagery. The first is traditional photo-interpretation, which requires a human to examine an image and extract features such as roads, water bodies, vegetation types and building structures. This is usually accomplished on-screen via heads-up digitizing.

An important rule of thumb is that the smallest item that can be reliably and consistently differentiated as a discrete feature during image classification is about 3 x 3 pixels in size. For example, if the requirement is to map land cover types across a state, imagery with a 30-meter spatial resolution will suffice. However, if the requirement is to map individual trees within a stand that have been infested with beetles, 30 meters is too coarse, and one-meter or sub-meter imagery is required.

Software developers are creating semi-automated classification schemes for specific features, such as railroad tracks, buildings and roads. This requires the software to recognize the same clues humans use to interpret images, including color, tone, texture, shape and context. A reliable way to automatically classify man-made features hasn’t yet been perfected though, and may require human intervention to edit the results. However, image processing software is continuing to improve in the realm of automated feature extraction, largely driven by the increasing number of multispectral digital sensors in use.

Bringing It All Together
Imagery’s utility continues to increase as improved all-digital acquisition and processing systems are introduced. Therefore, it is more important than ever for geospatial data users to coordinate imagery purchases with other divisions and departments within their organizations to determine if financial resources can be pooled to buy a data set that serves the needs of multiple users.

Although the imagery itself can be used as a map background, added value is gained when the imagery is converted into different classification layers that can be used as key inputs in geospatial models. One of the most beneficial applications of geospatial technology is conducting “what-if” modeling scenarios to analyze cause-and-effect relationships before implementing any action on the ground. Such applications turn ordinary image data into powerful information.

Publisher’s Note: For a complete overview of available digital airborne cameras, see “Digital Airborne Cameras—Choosing the Right Tool for the Job,” Earth Imaging Journal, March/April 2005 (http://www.eijournal.com/DigitalCameras.asp).
 

  See  Featured Images
 
  Subscribe to Earth Imaging Journal

  Go to Home Page