What is Hyperspectral Imaging?
A normal visible camera can see colours from red to blue and all the colours in between (like a rainbow spectrum). So what other possible information is there to be gathered? The answer is plenty.
Hyperspectral imaging, also known as imaging spectroscopy, can be used to detect whether vegetation is alive or dying even if the vegetation appears green and healthy to the naked eye. Hyperspectral imaging can also be used to classify different rock types in aerial imaging and much more.
Hyperspectral cameras can be a hard concept to grasp. At first the concept may not be obvious as our eyes only have three colour receptors (red, green, blue) yet a hyperspectral camera would be the equivalent of having hundreds of different colour receptors.
What is Hyperspectral Imaging?
Before explaining what a hyperspectral camera is we must step back to how monochrome and colour cameras work. A monochrome camera one has one "colour" sensor for each pixel which spans the full spectrum of visible light approximately 450 to 750nm. The sensor detects the presence of red, green, or blue light but it cannot differentiate between red or blue, we simply know how bright it is within the band of 450-750nm.
A colour camera has three overlapping "colour" sensors for each pixel. For our hypothetical colour camera will assume 450-495nm for blue, 495-570nm for green and 570-750nm for red the detection range of each sensor within a pixel.
If the wavelength of light landing on the sensors is exactly half way between two sensors (say 495nm) then we will see the colour light blue and detect two equal raw values for green and blue. Now say we shine 475nm and 530nm wavelength light on the same sensors, lets say the two intensities of wavelengths are the same. The blue sensor will only detect the 475nm wavelength, the green sensor will only detect the 530nm wavelength. They will appear the same magnitude and the colour will appear to be light blue, the exact same colour that we saw earlier. This is where hyperspectral cameras come in.
Figure 1: Colour camera sensor levels given certain wavelengths of light.
Figure 2: Colour camera sensor levels given a different set of wavelengths of light. Notice that the missing frequencies cannot be determined from the cameras sensors as they average out to simply give intensities of light on each sensor.
Hyperspectral cameras have the equivalent of hundreds of very narrow wavelength sensors which overlap to cover the full spectrum of visible light. In the case above, there will be one sensor which detects 495nm light, another sensor which detects 475nm light and another sensor which detects 530nm light. Each sensor detects it own narrow band of light. There is a huge array of hidden information contained in light that our eyes cannot see, just as a normal camera cannot "see" the hidden information. In reality a hyperspectral camera uses special optics to achieve the equivalent of having hundreds of sensors.
Figure 3: Hyperspectral camera sensor levels given certain wavelengths of light. Notice that the exact frequencies of missing light can be identified easily.
How Can Hyperspectral Imaging be used?
Some objects have spectral signatures (or 'fingerprints') that can be detected using a hyperspectral camera. Certain wavelengths can be absorbed, reflected, or emitted depending on the object. These signatures can be used to generate a software or hardware based filter to detect an object. There are many known applications for hyperspectral imaging, and many more applications that are being discovered every day.
Hyperspectral cameras can be used in R&D applications where the signature of the object being detected is not known. The results from the hyperspectral camera can then be used to select an appropriate multispectral camera which saves cost and simplifies image processing.
IMC has in house expertise collecting and analysing hyperspectral imaging data. We have extensive knowledge of the different existing applications. We have demo hyperspectral cameras available and can assist in performing trials to prove conceptual designs.
(Title Image By Aappo, CC BY-SA 3.0, https://commons.wikimedia.org/w/index.php?curid=17686740)