RICS Draft Global Guidance Note: Earth observation and aerial surveys, 6th edition

RICS Draft Guidance Note: Earth observation and aerial surveys, 6th edition

6 Hyperspectral, multispectral and thermal imaging sensors

Aerial survey sensors operating in the non-visible parts of the electromagnetic spectrum offer a rich source of data from which to extract information through sophisticated spectral analysis. When using these sensors, the emphasis tends to be on identification and the condition of features, rather than their absolute position. These sensors broadly fall into three categories: thermal, multispectral and hyperspectral. In recent years, the use of UAVs for precision agriculture has been an important factor in driving the development of small format multispectral and hyperspectral cameras.

Thermal imaging

Thermal cameras operate in the infrared part of the spectrum at wavelengths of 4-12µ microns. They can sense heat energy emitted, reflected or transmitted by an object. A good quality, well-calibrated sensor can have a measurement accuracy of +/- 1C. They have found applications in the power industry and in housing energy efficiency surveys.

Multispectral imaging

Multispectral cameras typically sense in between three and ten separate bands, depending on the application of the data. Band-pass filters are used to filter out unwanted frequencies. A common configuration is the use of three bands (visible green, visible red and a near infrared band), which finds applications in determining plant and soil water saturation.

The processing of multispectral data can be very complex and requires the specialist knowledge of a spectral analyst.

Hyperspectral imaging

Hyperspectral sensors have the broadest spectral range of all non-visible sensors. A high-quality instrument can typically sense from 380 to 2,500nm, from the visible near infrared (VNIR) to the short-wave infrared (SWIR) parts of the electromagnetic spectrum.

These instruments are also characterised by the large number of spectral bands that they can use to separate the data, typically more than 500. With such a rich data source, the applications are bespoke, numerous and varied, including:

  • the detection of invasive plant and insect species
  • precision agriculture and
  • the detection of minerals.

As for multispectral sensors, the processing and manipulation of this data requires the specialist knowledge of a spectral analyst.

6.1 Key considerations

6.1.1 GSD

GSD is a key metric specified by the contractor when commissioning hyperspectral, multispectral or thermal imaging projects. As for aerial imagery, the GSD of any hyperspectral, multispectral or thermal imaging mission is determined by the flying altitude, the dimensions of the camera sensor chip and the lens. However, when using hyperspectral, multispectral and thermal imaging sensors, improving the spatial and spectral resolution can increase the amount of noise in the signal. It is therefore not uncommon for GSDs of 0.5m or even 1m to be used, with the emphasis on feature identification and condition.

6.1.2 Directly geo-referencing hyperspectral, multispectral and thermal imagery

Hyperspectral, multispectral and thermal imaging sensors are passive sensors, relying on radiation from the sun reaching the camera head, or in the case of a thermal sensor, being emitted from the target. Both push broom and frame sensors are available, although push broom sensors are becoming less common.

They do rely on an accurate GNSS/IMU navigation system to directly geo-reference the imagery, similar to the systems described in sections 4.1.5 for aerial photography and 5.1.3 for LiDAR sensors.

It may be necessary to document the sensor models used to transform coordinates on the sensor device to coordinates on the earth's surface. Should a requirement arise to document the sensor models, the following specifications may be used to describe them:

  • OGC 12-000, OGC® SensorML: Model and XML Encoding Standard and
  • OGC 17-011r2, JSON Encoding Rules SWE Common/SensorML.

6.1.3 Calibration

Hyperspectral, multispectral and thermal imaging sensors are sensitive precision instruments. The quality of the data captured is very dependent on ensuring that the instruments are calibrated regularly according to the instrument manufacturer's guidelines.

As with aerial photographic cameras and LiDAR sensors, a calibration flight is necessary every time the instrument is installed in a new aircraft to establish the relationship between the sensor and the aircraft coordinate system.

6.2 Flying and coverage

6.2.1 Flight lines and overlap

As with LiDAR imagery, professional flight planning software is used by contractors.

Flight line planning for hyperspectral, multispectral or thermal imaging sensors follow the same principles as for vertical aerial photography detailed in section 4.2.1.

Overlaps between flight lines usually at between 15% and 35% are necessary to ensure that there are no gaps.

6.2.2 Flight times

Thermal imagery does not rely on the capture of the visible parts of the electromagnetic spectrum, and therefore it can be captured at night. Indeed, this is recommended, because air temperatures are cooler and more stable at night, meaning differences in temperature are easier to detect. Thermal imagery is frequently commissioned during the winter period for the same reasons.

Multispectral imagery is frequently commissioned for the study of vegetation health and land use cover and as such it is usually flown during the main part of the flying season when the vegetation is in full bloom.

Hyperspectral imaging sensors require the best conditions possible to detect enough energy to produce good quality results. This usually means operation with a high solar angle in bright sunshine and cloudless skies. The presence of a significant amount of moisture in the air will affect the returns in the SWIR part of the electromagnetic spectrum.

6.2.3 Acceptable quality limits

The following list is intended to act as a set of AQLs to provide guidance on the subjective topic of image quality. The client and contractor should work closely together to ensure a mutually acceptable result.

  • Geospatial and spectral resolution specifications should be met.
  • Full coverage should be achieved.
  • There should be a good match between flight runs and adjacent images.
  • Hyperspectral, multispectral and thermal sensors should only be flown in good conditions, in the absence of rain, cloud, atmospheric haze, snow and flooding.
  • Hyperspectral sensors are particularly sensitive to moisture in the atmosphere and should only be flown in bright sunshine and cloudless skies.

The intended use of the data may impose limitations upon times of flying. See project constraints in section 3.3.

6.3 Hyperspectral, multispectral and thermal imagery accuracy and resolution table

The emphasis of aerial surveys operating in the non-visible part of the electromagnetic spectrum has been on feature identification and condition.

Table 7 was prepared from first principles focusing on the achievable spectral resolutions.


Height AGL

Achievable resolution - GSD (m)
























Fixed wing






Fixed wing






Fixed wing






Fixed wing






Table 7: Achievable resolution values for thermal, multispectral and hyperspectral imagery

The UAV flying altitude of 400ft represents the highest altitude at which a UAV can be operated in the UK without the approval of an operational safety case.

6.4 Hyperspectral, multispectral and thermal imagery products

6.4.1 Thermal imagery

The most common product created from thermal imagery is an orthorectified mosaic created from the individually captured thermal images. In a mosaic, the individual pixels have values that reflect the relative temperatures across the AOI. The orthorectification process ensures that accurate measurements are possible from the imagery.

Thermal imagery has found applications in monitoring energy usage over wide areas and pinpointing inefficient assets such as individual buildings and pipelines.

6.4.2 Multispectral imagery

The simplest products created from multispectral imagery are three- and four-band orthorectified images. Colour infrared (CIR) photography combines the red and green visible bands with the infrared part of the spectrum. Intense reds in this imagery are associated with healthy vegetation showing fast growth rates. Four-band imagery - combining the visible red, green and blue bands with the near infrared (NIR) band - has all the benefits of traditional imagery, with the additional advantage that the NIR band can be used for vegetation studies.

Products created from ten-band multispectral cameras are much more complex to create. The successful analysis of this data relies on the creation of a spectral signature of the landscape feature in the terrain that the user wishes to detect.

A spectral signature is the variation in the electromagnetic reflectance of a homogeneous target across several different wavelengths. Spectral signatures are unique responses for each land cover classification, such as sand, roads, cereals, grassland, moorland, or forestry. They are created using a combination of multispectral observations from the camera and observations on the ground, known as 'ground truth' observations. Using a predefined spectral signature, an automatic classification is run on the imagery to identify the target object, such as the area of cereals under cultivation.

6.4.3 Hyperspectral imagery

The principles behind hyperspectral imagery classification are the same as for multispectral imagery in that a spectral signature is used to classify and extract the landscape feature under study.

Hyperspectral data is captured in a wide range of the electromagnetic spectrum (2,120nm) that is split into over 500 individual spectral bands; this improves the precision and the number of data processing possibilities. However, a lot more effort and attention are required when completing the 'ground truth' observations, which should include handheld spectrometer observations.

In the more traditional area of vegetation analysis, hyperspectral imaging has the potential of not just being able to identify cereals but also to differentiate between wheat and barley, for example.

Hyperspectral imaging has found applications in:

  • invasive species detection
  • forestry inventories
  • the estimation of soil water content
  • geological applications and
  • mineral exploration.