PTB >> PTB Insights: Superframing Increases the Temperature
Measurement Range of IR Cameras

Infrared cameras are versatile tools for measuring the temperature of objects in a scene without requiring physical contact. In addition, one can measure a temperature value for every pixel in the image, instead of just a single spot on the surface as with contact thermometers. But sometimes the objects in the scene under inspection are so hot that the camera output saturates, giving you an image that photographers would describe as “blown out.”

Infrared cameras, like any other camera, do not have infinite dynamic range. When an object in the scene appears blown out then two undesirable things happen: image details are lost and any temperature measurements in that part of the scene are no longer valid. This saturation problem, particularly acute with midwave infrared (MWIR) cameras because of the high thermal contrast in the MWIR waveband, can be addressed with a technique scientists and engineers in the military range community call superframing.

These users frequently encounter rapidly changing, infrared scenes with very large temperature ranges, such as rocket launches. Here the rocket body itself can be at sub-zero temperatures at the same time that the rocket exhaust plume has a temperature of many thousands of degrees.

MWIR Cameras
MWIR cameras operate in the 3-5-micron waveband, and are often designed around indium antimonide (InSb) sensors. These sensitive devices are capable of detecting temperature changes as small as 0.02°C under the right conditions. One can control the sensitivity of the camera by varying the exposure time. As with conventional cameras, the longer the exposure time, the higher the sensitivity.

Operating the camera at a very high sensitivity restricts the range of temperatures one can measure accurately because hot objects are so bright that they exceed the dynamic range of the camera. For example, a typical InSb camera with f/2.3 optics is operated at an exposure time of 2 ms when used to view a room-temperature scene with people in it (i.e., a scene with temperatures between about 20°C and 40°C). This is considered a long exposure time for an InSb camera and the system has lots of sensitivity. However, under these conditions the camera will saturate when viewing a target at a temperature of roughly 50°C or greater. One sacrifices temperature range for sensitivity.

If the scene contains both people and a hot machine (e.g., with surface temperatures of 200°C) that need to be simultaneously measured, the exposure time of the camera would then need to be reduced to a much shorter time, like 30 µs, so that the 200°C object in the image no longer saturates the camera’s sensor. But increasing the temperature dynamic range in this manner will result in the loss of the ability to measure subtle changes in temperature of the cooler parts of the scene, like the people. The upshot of all of this is that there is no single exposure time that allows for optimal measurement of all the objects in such a scene with an MWIR camera.

Figure 1 illustrates this effect with two images of a Beechcraft King Air twin propeller airplane taken at 2 ms and 30 µs. These images were taken with a very high performance MWIR camera system running at 90 frames/s at the full frame size of 640×512 pixels. The two images are separated by a short interval of time (about 40 ms), meaning that the scene does not change very much — the propeller movement is barely perceptible. This is possible because the system supports exposure-time cycling, which is to say that exposure times are changed on a frame-by-frame basis for four preset exposure times (typically), and then the cycle is repeated.

The 2-ms image gives excellent contrast for nearly every portion of the scene except for the aircraft’s exhaust system, which is so hot that it saturates that part of the image. Conversely, the 30-µs image shows the exhaust system very clearly without saturation, but the rest of the scene is too cold to see clearly above the system noise floor. Combining these two images with the right algorithm enables imagery that is both high in contrast and wide in temperature range — the best of both worlds. This superframing technique has already been implemented in commercially available infrared camera systems.

Superframing
Superframing typically involves taking a set of four images (subframes) of the scene at progressively shorter exposure times in rapid succession, then repeating the cycle of subframes. The subframes in a given cycle are then combined in post-processing to yield single images, or superframes, with greatly increased temperature dynamic range. The superframes are then combined into image sequences known as a “supermovie.” If the camera system runs at 100 frames/s and four subframes are used in a cycle, then the resulting supermovie has an effective frame rate of 100/4 or 25 frames/s after post-processing. This frame rate is sufficient for most applications.

Figure 2 shows a single superframe image taken by a MWIR camera system of the Beechcraft airplane. This superframe is part of a 200 frame supermovie that shows the start-up of the aircraft engines. The full temperature range of all the points in this scene is easily spanned without saturation and without sacrificing thermal contrast. The exposure times of the four subframes used to form this superframe image are 2 ms, 0.5 ms, 125 µs, and 30 µs, respectively. Figure 1 shows the first and fourth subframes of this superframe. There is very little apparent movement in the scene in between successive subframes because of the system’s high frame rate. Thus, the resulting superframes show crisp, clean edges, even along the moving edges of the propeller blade and the borders of the rapidly fluctuating exhaust plume.

The superframing technique is built on two enabling technological developments that have emerged in the commercial marketplace only in the last few years: One is the emergence of commercially available infrared cameras with large arrays (320×256 or 640×512 pixels) and high frame rates (up to 100 frames/s at 640×512 pixels or 346 frames/s at 320×256 pixels) necessary to produce superframes. The other is the availability of inexpensive consumergrade computers equipped with fast PCI buses, gigabytes of high-speed RAM, and disc drives with hundreds or thousands of gigabytes of storage capacity. This sort of computer is needed to swallow the enormous amounts of data (64 MB/s) coming from the infrared camera when operating at high frame rates.

Pixel Values
How does one calculate the temperature from the pixel values in each image? The camera system is typically calibrated for each exposure time in units of infrared radiance, which has the dimensions of watts/cm²-sr, using laboratory blackbody sources at precise temperatures. If greater precision in measurement is required, the operator can take pains to maintain the lens temperature in the field at the same temperature that it was during calibration. This prevents any offsets in the radiance measurement due to changes in the radiance of the lens itself.

When the subframes are collapsed into superframes during post-processing, every pixel in the resulting superframe is expressed in radiance units, rather than digital counts. The radiance value and the emissivity of the target can then be used to calculate the temperature of each pixel in the image, although many applications, such as signature analysis, require only that radiance be determined. The emissivity of a surface is the ratio of the in-band radiance emitted by the surface to the in-band radiance of a perfect blackbody at the same temperature.

Fortunately, many common objects (e.g., people, animals, vehicles, vegetation, and terrain) have emissivities close to unity in the MWIR band, making it straightforward to determine surface temperatures. In an outdoor environment especially, the background is typically quite cold relative to many common targets of interest. This means that it is often not necessary to compensate for reflections of infrared light from other sources in the vicinity of the target to get accurate surface temperature measurements.

This article was written by Austin Richards, research scientist at FLIR Systems, Indigo Operations, Goleta, California. Richards holds a Ph.D. in physics from the University of California, Berkeley. He is the author of “Alien Vision: Exploring the Electromagnetic Spectrum with Imaging Technology” as well as numerous articles on infrared imaging technology. For information, call (805) 964-9797, or visit www.flirthermography.com/superframing.