HYPERSPECTRAL VS MULTISPECTRAL CAMERAS: UNDERSTANDING ADVANTAGES AND LIMITATIONS IN SPECTRAL IMAGING
The main difference between multispectral and hyperspectral cameras is the number of bands they record and how narrow the bands are (i.e., the spectral resolution).
A standard definition states that hyperspectral cameras have more than 100 bands, whereas multispectral ones have fewer. But this definition does not consider the width of the spectral range or the sampling. That means that if a camera covers the spectral range 400 – 600 nm with 50 bands, it would not be hyperspectral, whereas if it covered 400 – 800 nm with the same sampling (meaning this time 100 bands), it would be hyperspectral.
We find this discrepancy irrelevant and prefer to speak about spectral resolution (FWHM, Full Width Half Maximum*), highlighting the ability of a camera to separate two consecutive spectral peaks from each other.
Hyperspectral vs. Multispectral data
Hyperspectral imaging involves capturing and analyzing data from a large number of narrow, contiguous bands across the electromagnetic spectrum, resulting in a high-resolution spectrum for each pixel in the image. As a result, a hyperspectral camera provides smooth spectra. The spectra provided by multispectral cameras are more like stairs or saw teeth without the ability to depict acute spectral signatures.
Hyperspectral imaging provides more detailed data than multispectral imaging, allowing more specific analysis and accurate identification of materials and substances. Multispectral imaging may not be able to distinguish between closely related materials due to the limited spectral resolution.
For most of the multispectral cameras on the market, the spectral range is limited to 400 – 1000 nm and the typical number of bands is between 4 – 5. Those are crucial limitations for many applications.
To illustrate the advantages of a hyperspectral camera compared to a multispectral one, we studied the sorting of shells among almonds. It is a typical application that requires highly accurate identification between materials that look very similar.
Advantages of hyperspectral vs. multispectral camera based on the spectral range
The reflection, absorption and emission features form spectra that are closely related to the molecular composition of the material being examined. Table 1. is well known and it highlights the spectral bands in which each of the most common molecules has its electromagnetic resonance overtone.
Table 1. Molecular electromagnetics resonance overtones
As the table shows, the spectral range of 700 – 2500 nm is necessary for many applications. Especially for those related to food quality assessment and plastics sorting, the spectral range of 1100 – 1700 nm is mandatory. Multispectral cameras limited to 400 – 1000 nm are not applicable for those applications.
Our article about Hyperspectral imaging vs. RGB cameras shows that RGB cameras do not perform well in sorting nuts and pistachios. The Specim FX10 provided better results, and the Specim FX17 had the most accurate sorting capability.
Advantages of hyperspectral vs. multispectral camera based on the number of spectral channels (bands)
Considering the previous point, we compared almonds and shells with Specim FX17 camera data. We covered the spectral range of 900 – 1700 nm in the first data set with 224 bands. In the second data set, we used only 28 binned bands (i.e., merging consecutive spectral bands) to simulate a multispectral camera.
As can be seen in Figures 1. and 2., spectra related to the 224 bands dataset are much smoother than those depicted with only 28 bands. From the hypersepctral data, we can also pick the small but crucial spectral differences that enable us to separate the almonds from the shells.
In other words, because the number of spectral channels captured is limited to 28, the spectral information required to tell the difference between the almond and shell is lost.
Figure 1. spectra of shell and almond depicted by 224 and 28 bands.
Figure 2. Specim FX17 spectrum of almond with 224 and 28 bands. Purple circles highlight the spectral differences due to the presence of oil in almonds but not in the shell.
In addition, some pre-processing methods are not suitable for multispectral data. For instance, derivative or smoother such as Savitzky-Golay, require continuous spectra to perform well, which are not provided by multispectral sensors.
As mentioned above, we built two models to illustrate these points related to the ability to depict acute spectral features. The hyperspectral model covering 224 bands is more accurate than the multispectral model covering only 28 bands. With the hyperspectral data, the edge effect disappears, and we can classify even the smallest pieces of shell correctly.
Figure 3. RGB, multispectral (28 bands), and hyperspectral (224 bands) images of almonds and shells (green is almond and blue is shell).
We selected 28 spectral bands in this study to simulate a multispectral camera. However, typical multispectral cameras can have notably fewer spectral bands, making their ability to depict fine spectral features even lower.
What are spectral channels (bands)?
Spectral channels are distinct wavelength ranges within the electromagnetic spectrum that sensors in multispectral or hyperspectral imaging systems capture. Each channel represents a specific band of wavelengths, and the collection of these channels helps to identify objects or materials based on their unique spectral signatures.
How to choose between hyperspectral imaging vs. multispectral?
Both multispectral and hyperspectral imaging are techniques used to capture and analyze the electromagnetic spectrum for various research, industry, and remote sensing applications. Choosing between these two complementary technologies comes down to the application requirements and the level of data already existing.
For applications that require more spectral bands and higher spectral resolution than multispectral imaging can provide, a hyperspectral camera is the natural solution.
For applications where the entire spectral range is unnecessary, custom-made multispectral cameras with well-chosen bands can perform equally well as hyperspectral cameras. However, this requires the user to know the selected number of spectral bands that are necessary for the inspection or analysis. If the user does not know the spectral requirements of an application or they are very complex, a hyperspectral camera provides an ideal solution to gather the data for analysis.
The hyperspectral camera also offers more flexibility as the user can later upgrade the machine to sort new contaminants or materials. With the Specim FX cameras, users can freely select relevant bands. In practice, the Specim FX hyperspectral cameras can transform into multispectral cameras, whereas a multispectral camera could never become hyperspectral.
Lastly, one aspect to consider when choosing between hyperspectral and multispectral cameras is the price. Typically hyperspectral cameras are more expensive and require more processing power than multispectral cameras. However, this is not always the case, especially if the multispectral camera needs to be custom-made with a relatively large number of bands.
*The full width half maximum defines the spectral resolution of a system, more precisely, in this context, its ability to separate spectral peaks from each other.