Image Analysis

Imaging helps us detect many different kinds of phenomena or events, from tracking storms to detecting troop movements and rocket launches. To use images, we need first to detect the image, classify the image (and determine its relevance to our purpose), interpret the image, and, finally, decide what our response to this image should be.

The number of surveillance systems, of all types, in use in the United States is truly staggering. These systems produce vast quantities of images, quantities that are orders of magnitude above the number we can analyze and interpret. Culling the trivial and retaining the significant comprise the biggest challenge to using images in biosurveillance.

Reliable automated image recognition and classification comprise a field still in its infancy; accurately interpreting images and communicating their significance are tasks for highly trained professional analysts. A competent image analyst boasts a good eye, relevant domain knowledge, and an ability to accurately translate what he/she is seeing into information that decision makers can use. Further, analysts specialize within domains and professions. For example, a radiologist is a highly skilled image analyst who understands how his instruments (x-ray, fluoroscope, CT scanner, MRI scanner) prepare images, the scale of those images, correlations between the images and the human body systems they portray, and the needs of the physician who requested the imaging. He can also aim the instruments so that the resulting images are more likely to reveal details relevant to the patient's case, such as a tumor in the spinal cord or an abscess in the liver.

Similarly, a government intelligence analyst understands how a reconnaissance satellite or aircraft's camera, radar or other sensor prepares images, what scale to use while viewing them, the characteristics of objects likely to appear in those images, and what each type of object conveys in terms of a target country's military or economic capabilities. The analyst can order an adjustment to the sensor's aim in order to optimize the value of the resulting images, for example, adjusting an orbital track and camera angle to provide greater detail about a missile installation or ground vegetation.

However, a radiologist's domain knowledge will not allow her to competently conclude whether a shape photographed from 200 miles away is part of an oil refinery or a facility for manufacturing biological agents, just as a military image analyst cannot competently diagnose a bowel obstruction by looking at an x-ray picture.

Was this article helpful?

0 0

Post a comment