DRI Explained: Thermal Infrared Camera Ratings

Infiniti Whitepaper

If you’ve done any research comparing thermal cameras, you’re likely to have come across the term “DRI”, which is often used to compare performance between thermal imaging cameras. DRI stands for Detection, Recognition and Identification, however it’s important to understand how those terms are defined, as they’re likely to mislead many customers with unrealistic expectations.

Note: As for more common color surveillance cameras (visible/NIR), there is another standard that sounds similar to DRI, but is different in its definitions. This standard is called DORI, which stands for Detection, Observation, Recognition and Identification. While most of the terms share the same words as DRI, it’s important to recognize that their specifications are quite different.

Detection Rating (DRI) for Thermal Infrared Cameras
Based on the DRI standards, detection means that you will be able to see the target, however it will be little more than a speck. There is a 50% chance that the target will be visible on at least two pixels, and and a skilled operator should be able to tell if the target may be something of suspicion.
Read More
PPM (Pixels Per Meter)
The amount of potential detail that a camera offers at a given distance. A higher PPM value means that the image definition is more detailed.
Read More
FLIR (Forward Looking Infrared)
Refers to the technology used to create a thermal infrared image of a scene without having to “scan” the scene with a moving sensor. It is also the name of a thermal imaging camera manufacturer (FLIR Systems).
Read More

What is DRI?

DRI is a universally accepted set of standards that attempts to provide a means of measuring the distance at which a thermal sensor can produce an image of a specific target. These standards were initially developed by the US Army in the 1950s and unfortunately have not been updated much to account for newer technology with larger theoretical ranges that are less likely to be achieved in the real world.

The model is based on a 50% probability of achieving the following objectives, under ideal conditions. Please read each definition as they may not match what you would expect from reading the terms alone.


Detection refers to the distance at which a target initially appears in the image. This “target” is something out of the ordinary that is warmer or cooler than the ambient environment. Specifically, it will be visible on at least two pixels, so there will not be enough information to confirm what the target is at this distance, just that something is there.


Contrary to what might be expected, recognition does not mean that you can recognize an individual. Recognition refers to the distance at which you can determine an object’s class (is it human, animal or vehicle).


Identification refers to the distance at which you can differentiate between objects within a class. For example, identifying the type of vehicle (truck, SUV, or car) or whether the human is a soldier or civilian.

Note that these distance measurements are based on a 50% probability and do not take any atmospheric conditions into consideration. Weather is almost never ideal so in real use these distances are almost always shorter than specified.

An outdated specification

So the terms detection, recognition, and identification can be misleading, especially to end users who do not have a military or electro-optics background. To make matters worse, the original 1950s specifications were based on old sensors and screen display technologies. The increasing resolution of thermal sensors has shrunk the size of the DRI areas to tiny specks of white on the screen.

For example, our standard uncooled thermal sensors have a resolution of 640×480, which is over 300,000 pixels. Human “detection” only requires the target to be visible on 3–4 of those pixels. This is an extraordinarily small portion of the screen that can easily go unnoticed by the human eye. In fact, if this page were the size of your video feed, the area required for a human detection rating would be equivalent to the size of this rectangle:   

This area becomes even smaller when you consider our HD sensors, which have resolutions over four times larger at 1280×1024 (over 1.3 million pixels).

Even when magnified, the amount of detail visible at the DRI distances is not as high as one might expect, as seen in the chart below.

Chart of DRI detail levels for human and vehicle

Note that while it may seem impossible to tell that a human or vehicle is present in some of these images, it’s often the movement of pixels in an area that makes it more obvious that an object is present. In addition, these standards were developed with trained professionals in mind, who will be able to spot activity much easier than a civilian.

Alternate Methods

The examples above show simulated detail levels at the distances Infiniti uses for DRI, but it’s important to know that different manufacturers may use different methods to determine their DRI numbers.

For example, L3 and FLIR have cameras that are almost identical in specification and should be within 5% of each other, but due to different assumptions on the conditions and atmosphere, one is rated for 21km of detection while the other is rated for 52km. Clearly these ratings are too arbitrary for proper comparison which is why for the sake of transparency and simplicity, Infiniti bases our DRI numbers on pixels on target, however pixels on target is not the only method available.


The US Army has an updated thermal ranging model called NVThermIP, which is more mathematically sophisticated in order to yield more accurate results. It requires inputting a variety of values that specify the camera lens, detector, framing and sampling electronics, signal processing electronics, viewing display, atmosphere, target and task. The problem is not only are these a lot of inputs, but quite often those values aren’t even known. In addition, there are so many variables and equations at play that it’s often unclear what’s going on.

Pixels on Target (PPM)

Used by many manufacturers as a simpler alternative to NVThermIP, this method is done by calculating the number of pixels needed across the critical dimension of the target and then converting that value to the required pixels per meter. The sensor and lens measurements then allow us to easily calculate the pixels per meter performance of the camera across a full range of distances and calculate the distance where each level of detail is achieved.

Of course a drawback to this method is that variables like weather and atmosphere are not included in the calculation, but those values are rarely accurately known anyways.

At Infiniti, we calculate the critical dimension by using the square root of the target height multiplied by its width, resulting in a critical dimension measurement for a human at 0.95m and a military vehicle at 2.2m. Using the Johnson Criteria of one line pair for detection, 3 line pairs for recognition and 6 line pairs for identification, and assuming one line pair is equal to two pixels, the required DRI values for PPM are as shown in the chart below.

Human Detection

2.1 ppm

Human Recognition

6.3 ppm

Human Identification

12.6 ppm

Vehicle Detection

0.9 ppm

Vehicle Recognition

2.7 ppm

Vehicle Identification

5.5 ppm