Airborne Hyperspectral Imaging vs. Other Remote Sensing Technologies
By Casey Smith, CTO - January 24, 2024
Airborne hyperspectral imaging systems, mounted on UAVs or
piloted aircraft, collect detailed spectral information from large areas, like
forests, crops, waterways, or mining operations. The hyperspectral data are
useful for numerous applications, including precision farming, environmental
monitoring, forestry, geological and mineral exploration, and archaeology.
Hyperspectral imaging is one of several different remote
sensing technologies available, some of which are complementary, some of which
have competing capabilities. Here, we briefly review a few of the best-known remote
sensing technologies.
LiDAR
Light Detection and Ranging (LiDAR) is a remote
sensing technology that utilizes laser illumination. Thus, it is an active
remote sensing technology that requires its own light source. There are two
common LiDAR approaches. Sensing LiDARs use two or more laser
wavelengths, or dither the laser wavelength, to probe a specific spectral
feature such as an absorption line. These sensors provide excellent
sensitivity, but they are typically only capable of sensing one entity at a
time, such as water vapor, carbon-dioxide, or methane
gas.
The other common remote sensing LiDAR measures
distance from the sensor to objects on the ground. With this information,
detailed 3-dimensional maps can be generated, but these maps reveal little or
nothing about the composition of the objects. For example, the LiDAR generated
image shown in Figure 1 clearly shows the structure but not the composition of
the objects. Thus, this type of LiDAR is complementary to hyperspectral remote
sensing in that it provides shape while hyperspectral imaging enables one to
determine the composition – both may be important for certain applications.
Figure 1: Image of a 3D model of London generated from airborne LiDAR data. Image courtesy Ephramac, CC BY-SA 4.0, via Wikimedia Commons
Thermal Remote Sensing
Unlike hyperspectral
remote sensing, which relies on reflected light, thermal remote sensing systems
measure the radiation emitted from objects themselves. Generally speaking, the
warmer the object, the more radiation it emits— consequently, the temperature
of the object can be determined.
Most thermal remote
sensing systems are sensitive to wavelengths in the Long Wave InfraRed (LWIR)
between 8 µm and 12 µm, a spectral range at which an object with even moderate
temperature emits radiation. The data output by a thermal remote sensing system
consists of a two-dimensional image with a single temperature value assigned to
each pixel. See Figure 2 for an example.
Since hyperspectral
imaging provides the spectral signature of the objects being scanned, which
allows one to determine the composition of the objects, thermal remote sensing
is generally complementary to hyperspectral remote sensing.
Figure 2: Thermal image taken from aerial inspection of solar panel modules. Note that damaged cells can be seen via small hot spots (nearly white color). Image courtesy Pix4d.
Hyperspectral Remote Sensing
Figure 3: Schematic of a hyperspectral datacube.
Airborne hyperspectral imaging uses solar illumination and thus is a passive remote sensing technology. Hyperspectral data can be visualized as a datacube—a 3-dimensional image with two spatial axes and a spectral axis, as shown at left.
The datacube provides
high-resolution spectral data with many tens or hundreds of spectral channels. Because
the spectra encompass a broad range of wavelengths, the data can be used to
identify many different types of materials and situations (e.g., plant species,
insect infestations, crop health, minerals, soil health, etc.) on a per-pixel
basis, as shown in the Figure 4, below.
Figure 4: Hyperspectral data of crop land obtained with a Pika L airborne system. The data are represented with false colors denoting spectral similarity. The data are overlaid atop Google Earth imagery of the same field.
Figure 5: Plot showing mean spectral data of the four locations identified in Figure 4, above.
Hyperspectral vs. Multispectral and RGB Remote Sensing
RGB images, like those you
take with a cell phone camera, provide three channels of data—red, green, and
blue. Multispectral images consist of 4 to 16 spectral channels, so they
provide more information than RGB images, but they don’t provide nearly as much
data as hyperspectral images. For example, the Pika L and the Pika IR-L,
Resonon’s two most popular hyperspectral cameras for airborne hyperspectral
applications, have 281 and 236 spectral channels, respectively. The differences
between RGB, multispectral, and hyperspectral data are shown schematically below,
with hyperspectral data most closely
representing nature.
Figure 6: Comparison between how a natural spectrum of light is represented with RGB Imaging, Multispectral Imaging, and with Hyperspectral Imaging.
The abundance of data provided
by hyperspectral imaging allows your remote sensing system to detect incredibly
subtle spectral differences between objects — differences that can be invisible
to other imaging modes, including the human eye. This fine spectral resolution
can increase detection sensitivity in some applications (e.g., allowing you to
detect and attend to crop stress earlier) and can enable applications that are
not possible with other imaging modes.
Review our extensive FAQ section for more information on Resonon’s airborne hyperspectral imaging systems.
