Image recognition technology is relatively young. To quote a timeworn cliché, our true state of the art in image recognition is the "Mark One eyeball,'' a joking reference to military weapon naming conventions.
Researchers are pursuing automated recognition via a number of techniques, such as pattern matching, use of "primitives,'' and others. The military has demonstrated its utility; in 1991, cruise missiles fired from aircraft and naval vessels employed image recognition technology during the final homing phase of their flight to strike their targets. One key to their success was the relative simplicity of the images that their seeker heads sought, such as a building window. Current weapons utilize improved software; the exact capability of that software is classified.
This simplicity can, potentially, be found in public health-related tasks as well. For example, a dead cow assumes a profile quite distinct from a sleeping cow, or one that is contentedly munching on alfalfa. If a computer program can be designed to recognize the first case, the latter two cases should provide more than enough variance to convince the program to reject them, given sufficient image resolution. A large group of dead cows would be spotted in the same way.
Humans could present a more difficult problem. Our bodies are smaller than cattle, and discrimination between death and sleep, or even death and any stationary pose, is more challenging. One way of making this distinction is by using infrared sensors. Infrared imaging, making use of the portion of the electromagnetic spectrum just above visible light, traces its practical origins to the discovery in the 1940s that exposing lead sulfide to heat energy reduces its electrical resistance, a property known as photoconductivity. Thus, a lead sulfide photocell generates an electric current when in the presence of a heat source (Westrum, 1999). The military took advantage of this beginning in the early 1950s.
By using a combination of optical and infrared photography, a surveillance analyst should be able to spot groups of human beings, and determine whether they are alive or dead. Other indirect means may include monitoring the level of radio traffic, which is currently easily accomplished by military satellites.
However, significant changes in the condition or activities of groups of humans are readily documented by satellite data; fully automated detection of desired signals, however, may require technologies that, if they exist, are classified. It is clear, however, that if satellites offering very high-resolution photography, coupled to appropriate enhancement and interpretive or detection software, were to be available to public health authorities, they could add to the U.S. ability to detect a significant disease outbreak or bioterrorism event.
Was this article helpful?