EO/IR, radar and sensor fusion explained

An aircraft's sensor list tells you what kinds of information it can collect, but it does not by itself establish how well the complete system performs. Visible cameras, thermal imagers and radar observe different physical properties. Combining them can make a picture more useful, provided the measurements and their limitations are handled correctly.

The useful distinction is between collecting a signal, interpreting it and establishing a dependable conclusion. A clear picture, a radar return and a software-generated label are different kinds of evidence.

EO: familiar imagery with familiar constraints

EO means electro-optical. In defence product descriptions, EO/IR commonly refers to a package containing visible-light and infrared imaging channels. FLIR's introductory explanation describes the combination in those terms. The visible channel gives an image resembling an ordinary photograph or video, which can make shapes, markings and the surrounding scene easier for people to interpret.

Visible cameras depend on light reaching the sensor. Lighting and visibility therefore matter. A statement that a payload provides high-definition video says little about the detail available in every real scene. Readers should look for the actual channel and configuration, rather than assuming all images from a multi-sensor payload have the same resolution or field of view.

IR: thermal information is a different view

IR means infrared. Thermal imagers turn detected infrared radiation into an image, often displayed using greyscale or a chosen colour palette. FLIR explains how thermal imaging can reveal differences when there is no visible illumination. The display colours are a presentation of measurements, not the natural colours of the objects.

Thermal imagery is not a universal ability to see through obstructions. FLIR's guidance explains that walls and trees block the direct view and that fog and rain can substantially reduce useful range. Performance depends on the conditions and the sensor. Avoid converting a general day/night capability into an unqualified all-weather claim.

Nor does every bright area provide a simple, exact temperature reading. FLIR's radiometry guidance describes the influence of emitted and reflected radiation and the surface's emissivity. A radiometric camera estimates temperature under assumptions and corrections; an attractive thermal image alone is not a calibrated measurement report.

Radar: a different measurement principle

An active radar transmits radio-frequency energy and measures returning signals. NASA's radar explanations show how the returned energy and its timing can support imaging. Because a radar supplies its own illumination, its operation does not depend on sunlight in the way an ordinary visible camera does.

Synthetic aperture radar, or SAR, combines measurements collected as the platform moves to produce detailed imagery. NASA describes this as a processing technique used from both aircraft and spacecraft. It is one radar capability, not a synonym for every radar.

Radar imagery also needs interpretation: bright and dark areas represent differences in returned energy, influenced by surface properties and viewing geometry. NASA notes that appropriate radar wavelengths can image through cloud and dust that obstruct visible and infrared instruments. That useful distinction does not imply that every radar mode identifies every object or that weather and installation have no consequences.

What sensor fusion adds

Sensor fusion combines information from different sources to produce a more coherent estimate or picture. The UK Ministry of Defence's explanation of SAPIENT separates fusion—correlating, associating and tracking information—from sensor management, which directs what sensors do. Simply placing two live feeds beside one another does not necessarily demonstrate that their information has been fused.

As an illustrative reading aid, imagine a maritime picture in which one display contains camera imagery and another contains radar observations. A fused presentation might associate observations thought to describe the same object. The association itself is an assessment, and its uncertainty matters. This example explains the concept; it does not claim a particular aircraft implements it in a particular way.

NIST's work on sensing and robotic navigation emphasises that imperfect or inconsistent information creates uncertainty. Additional sensors do not remove the need to measure it. Confidence in the combined output should rest on evaluation, including cases where the inputs disagree or information is incomplete.

Reading capability claims

Look for a named sensor configuration, the function being demonstrated and the evidence behind the claim. Keep detection, tracking, classification and identification separate: observing something is not the same as establishing exactly what it is.

Also distinguish offered features from integrated equipment and tested performance from general product descriptions. A useful profile explains what each sensor contributes, acknowledges unknowns and links its claims to sources. The strongest evidence concerns the complete configured system under stated conditions, with uncertainty still visible to the reader.

From a measurement to a useful result

A sensor produces measurements; software and people turn them into an interpretation. Those stages should remain visible in reporting. A label placed over an image is not another independent observation. Nor is a smoother display necessarily a more accurate estimate. Processing can make information easier to use while still inheriting limitations from the original observations.

For a fictional building-inspection example, a visible image might show the location of a roof panel while a thermal image highlights a different radiation pattern. Together they help an inspector decide what to examine further. Neither image alone establishes the cause. The useful outcome is a documented inspection finding with appropriate uncertainty, rather than a colourful picture presented as a complete diagnosis.

Resolution, field of view and context

A camera's pixel count is a property of its image format. It is not a guarantee of how much detail a particular subject occupies in the picture. Framing, optics, motion, lighting, transmission and processing all affect the delivered view. A close-up may provide detail while losing context; a broad view may show relationships while leaving small features unresolved. Treat those as different information products rather than assuming one headline number settles the comparison.

The same care applies to video. A demonstration edited into a short clip may omit pauses, unusable frames or operator interventions. Ask whether the published material represents a continuous task, selected highlights or a controlled demonstration. Without that information, describe what the material shows and avoid a general reliability claim.

Worked example: when two sources disagree

Imagine an invented survey in which a camera image and another sensor's location estimate appear inconsistent. There are several possible explanations: the observations could have been taken at different times, referred to different objects or used different coordinate assumptions. The disagreement is a reason to investigate provenance, not to choose whichever output looks more precise.

A useful fusion report would identify the inputs, their timing, the association being proposed and the uncertainty in the result. If the system cannot resolve the disagreement, that limitation should be visible to the person using it. This explains an evaluation question; it does not describe the internal design of a particular military payload.

Read the demonstration as carefully as the specification

Look for the named sensor, software baseline and test conditions. Was the claimed result detection, tracking, classification or identification? Was a human interpreting the output, and was the interpretation independently checked? How were ambiguous cases handled? A statement that a payload is compatible with a vehicle is weaker evidence than a documented integrated demonstration, which in turn may establish less than sustained use.

The photograph accompanying this guide is a historical aircraft image. It does not show the actual sensor output discussed in these examples. The comparison visual is an editorial explanation, not a simulated camera or radar image. Maintaining that distinction prevents illustrations from being mistaken for evidence of performance.

Suggest a correction to this entry