Contents
Thermography with drones - it's the application that counts!
The use of drones as a thermal imaging solution has developed rapidly in recent years. More and more often, professional users are deploying drones specifically for thermographic purposes. Thanks to the simple handling of the technology, its use has now also become child's play for the layman. Nevertheless, you should be aware of the most important boundary conditions, as not all of us are certified thermographers. We provide an overview of the theory and technology.
An image is created from a surface
Thermography is based on measurements of infrared radiation in the electromagnetic frequency spectrum. The intensity of emitted infrared radiation on the body's surface is recorded without contact using infrared detectors, colloquially known as "thermal imaging cameras". The measurement result is displayed in false colors, as the original wavelength of the infrared radiation is not visible to the human eye. The measured values are then displayed in grey or color value images, whereby each value is assigned to an intensity using a greyscale or color scale. If radiometric infrared measuring devices are used, a temperature can also be assigned to a measured value. However, this requires complex calibrations. The manufacturer FLIR identifies these devices with an "R" (short for radiometric) in the product description and they are correspondingly more expensive.
To keep sensors on drones light, small and compact, uncooled microbolometers are used. The built-in thermal imaging sensor absorbs incoming radiation in the infrared range (wavelength 8 to 13µm). The temperature change inside is evaluated. In science, cooled systems are used to detect the smallest temperature differences. The sensitivity of the sensor is considerably higher.
However, relative intensity measurements are sufficient for most applications in combination with drones. Knowledge of possible influencing factors on the measurement result is crucial.
What do I have to pay attention to when capturing images?
Every surface emits heat in the form of infrared radiation (emission). An infrared image is therefore always the sum of all the radiation hitting the lens or the sensor behind it. This is significantly influenced by the emission and reflection of the surroundings.
It must be possible to clearly distinguish the emission of an object from the emission of its surroundings. This is particularly important on warm, sunny days. Otherwise, objects would only be moderately distinguishable from their surroundings. Wind or rain also lead to heat convection on the surface and thus distort the measurement result.
If ambient radiation, such as the sun, hits an object, this is also reflected depending on the shape and color of the surface. Brighter or shiny surfaces, e.g. on metals, lead to increased reflection. Solar radiation can therefore exceed the emission of a surface due to strong reflections and cause misinterpretations.
This means that it must also be ensured that what I want to recognize can be recognized at all. Otherwise, images are obtained that are meaningless.
In order to find corresponding temperature differences during a measurement, it is important to pay attention to the right time of day when rescuing fawns, for example. The fawn that you want to recognize must have a body temperature that is sufficiently different from its surroundings so that it can be identified.
In general, it is also not advisable to take oblique images in thermography in order to avoid parallel projections. Abnormalities are interpreted or documented in the wrong way if the recording angle is disregarded. This is a particular challenge for thermography on curved surfaces. In order to achieve the best possible result, it is therefore advisable to consider the following correlations.
Relationship between distance and level of detail
The resolution describes the level of detail of the image capture, expressed in millimeters per pixel. The focal length, distance or altitude and number of pixels (megapixels) of the camera sensor have an influence on this value.
The greater the altitude or distance to the object, the larger the image section captured, but this reduces the level of detail in the image. The decisive factor for an optimum distance is the smallest distance that can still be detected. For each flight, the question must be asked as to whether anything at all can be detected with the selected measures. In order to be able to detect thermal bridges on buildings or faults on solar systems, these must either be well advanced or the flight altitude must be reduced accordingly.
For a comparison, three apple juice bottles at different temperatures were photographed from a distance of 5m. While the cooled bottle on the left in the picture still stands out very well from its surroundings, the heated bottle on the right is difficult to recognize. A low resolution makes the two warmer bottles disappear completely.
The flight altitude is particularly important when using a low-resolution bolometer for fawn rescue. In the worst case, you have to fly so low that the animal is more likely to be startled by the propeller noise than to be detected. It is better to use a higher-resolution thermal imaging camera to detect fawns from greater heights.
What else is important for the sensor?
In addition to the resolution (specified in px), the key performance characteristics of a thermal imaging sensor are
- Sensor size and focal length
- Sensitivity (specified in mK) and image optimization
- Refresh rate (specified in Hz)
- Quality of the lenses
Basically, the combination of the sensor size of the bolometer and the focal length of a lens results in the image angle. Apart from the distance, this determines the image section. The size of an image at the same distance depends only on the focal length.
With the same sensor size, long focal lengths act like a telephoto lens and short focal lengths like a wide-angle lens. The latter has the advantage of capturing a larger image section, but increasingly distorts the subject towards the edge. A person's perception of perspective corresponds to an image angle of 40° to 50°. A sensor/lens combination in this range therefore results in the least distortion in the image.
The sensitivity of a thermal image sensor indicates which temperature differences can be distinguished. The smaller the value, the more precisely deviations can be measured. For outdoor applications, less than 100mK is common. As only temperature changes are recorded in the sensor, the sensor is reset to reference values at regular intervals. This process is known as non-uniformity correction (NUC) and can be heard by the mechanical clicking sound.
The refresh rate and resolution of commercial infrared detectors are limited by application and export regulations. Manufacturers therefore often follow the American specifications and limit the resolution of the image sensor to a maximum of 640 x 512 pixels. Cameras with frame rates higher than 9 Hz may only be exported to authorized countries. Higher frame rates require an end user agreement. The low frame rate (9 frames per second) is particularly noticeable in moving images.
The lenses of a bolometer are not made of glass, but of a zinc compound or germanium. This is because glass allows infrared radiation to pass through only poorly or not at all. For this reason, it is not possible to see through the lens with the naked eye.
Opportunities and risks of thermography from the air
The opportunities for thermography from the air are obvious. Temperature distributions and hot spots can be recorded over a large area instead of at specific points. The method is also contactless and can be used from a great distance, even at night. This allows risks to be identified before they lead to damage.
However, the limitations must not be ignored. As only surfaces are measured, the measurement image is difficult to interpret without knowledge of the effects of emissions. Reflections can also cause considerable interference and the accuracy of a radiometric measurement is limited. In order to be able to make reliable statements, it must be ensured that what is to be detected can also be detected.