Detection

What is “Detection”?

Infrared surveillance has three terms that sometimes are used casually or incorrectly. But each has a precise definition, namely, that an observer has a 50% probability of correctly performing a particular task:

• Detection – The observer can tell “something” is within the field of view. That is, the observer can distinguish an object from its background, but cannot tell what it is. The “critical dimension” for detection is 2.0 ±0.5 pixels. In general, for humans this means the vertical dimension. For vehicles, it means the horizontal dimension. For comparison, consider that a typical computer display is 1920×1080 pixels. So, a “detected” object is very small.
• Recognition – The observer can recognize an object’s category, e.g., car, truck, or human. The critical dimension for recognition is 8.0 ±1.6 pixels.
• Identification – The observer can pick out identifying features, e.g., what kind of car, how a person is dressed, whether a person is carrying a rifle or a shovel. The critical dimension for identification is 13.0 ±3.0 pixels.

John Johnson, a scientist at the U.S. Army’s Night Vision & Electronic Sensors Directorate (NVESD), created these “Johnson criteria” in 1958 while working with image intensifiers. Subsequent work by Night Vision Labs resulted in more exact modeling that accounts for real-world factors.

Factors Affecting Detection

Actual performance of a surveillance system is affected by numerous variables. Some are environmental, some are intrinsic to the system design. These include:

• Sensor characteristics: pixel size, pixel pitch, array size, …
• Lens characteristics: focal length, aperture (F-number), field of view, …
• Camera height above ground level
• Air clarity: humidity; weather conditions such as rain, mist, or fog;
  presence of particulates such as dust or smoke, …
• Air turbulence and variations in density
• Target characteristics: size, temperature, motion, …
• Background characteristics: temperature, texture, amount of “clutter”, …