Hyperspectral imaging is the most comprehensive of the three common image processing technologies. The other two technologies are red-green-blue (RGB) imaging and multispectral imaging. All three are noninvasive and nondestructive and give engineers and scientists different ways to analyze objects.
RGB imaging can be quick and inexpensive to implement and provides basic information about an object. Multispectral imaging captures more nuances in the various wavelengths that make up the visible spectrum. Hyperspectral imaging captures detailed information about each pixel across a much wider spectrum, typically from 250 nanometers (nm) to 15,000 nm and thermal infrared (Figure 1).
Multispectral versus hyperspectral
Multispectral imaging is a refinement and extension of RGB imaging using more bands of spectrum. The resulting data can be sufficiently detailed to enable analysis of an object’s basic physical and chemical characteristics.
Hyperspectral imaging is different. It combines imaging with spectroscopy. With hyperspectral imaging, the spectrum of each pixel is captured in detail. It produces data about the spatial and spectral content in an image. Hyperspectral imaging supports detailed characterization of an object’s composition.
A typical hyperspectral camera captures hundreds of thousands of spectra to create a hyperspectral cube where two dimensions represent the spatial structure of an image, and the third dimension is the spectral content. In addition, the captured information is sufficiently detailed to be presented as a continuous representation and not in discrete spectral buckets like those used for RGB and multispectral imaging (Figure 2).
Commercial development of hyperspectral imaging has been enabled by advances in imaging, including the ability to quickly separate the reflected light from an object into its spectral components using techniques like spatial scanning, spectral scanning, snapshot imaging, and spatio-spectral scanning. The availability of high-resolution CMOS sensors that operate at video rates coupled with advances in high-performance image processing systems are also key factors.
What does a hyperspectral camera look like?
There are several ways to make a hyperspectral camera. One embodiment shown in Figure 3 starts with a lens that focuses an image of the object on a narrow slit that defines the imaging line. The narrowness of the list causes the light to diffract, and the collimating lens is used to align the beam and remove the effect of the diffraction. The light then passes through the prism-diffraction grating-prism structure that separates it into its spectral components. Finally, a lens focuses the resulting light onto a CMOS sensor (not shown) that produces the hyperdata cube. Also not shown is the mirror structure that is used to scan the scene and produce the individual line scans.
What is hyperspectral imaging good for?
Different materials can be identified by their spectral signature. While it began with multispectral imaging, the increased level of detail enabled by hyperspectral imaging has elevated that ability for remote identification and analysis of materials. Hyperspectral imaging can be implemented on production lines, aircraft, drones, and satellites. It’s used across a wide range of applications, including astronomy, agriculture, geology, biomedical imaging, environmental monitoring, and more.
Hyperspectral cameras are available from several vendors and have been optimized for specific uses, including:
- Hyperspectral imaging is used to assess the health and nutrient levels in crops enabling farmers to selectively apply fertilizers and insect control measures, reducing costs, and maximizing results.
- Art analysis. Checking the chemical content of the materials in an art object can ensure that they are consistent with the period and can support the dating of newly discovered works of art.
- Food processing. In the meat industry, hyperspectral imaging is used to measure the chemical composition to determine the protein content or identify bone, cartilage, and other materials. It’s also used for general quality control processes like determining the ripeness of fruits and vegetables.
- The ability to identify materials using hyperspectral imaging supports automated sorting and management of recycled items.
Summary
Hyperspectral imaging captures and processes information about an object across a wide range of wavelengths, often extending from 250 to 15,000 nm and thermal infrared. It produces much more nuanced information than multispectral imaging and enables a more detailed analysis of materials and processes. Commercial hyperspectral cameras have been designed for a range of specific applications.
References
Hyperspectral Applications and Case Studies, RESONON
Line Scan Hyperspectral Imaging Framework for Open Source Low-Cost Platforms, MDPI remote sensing
What is Hyperspectral Imaging?, NIREOS
What is Hyperspectral Imaging?, SPECIM
Why Isn’t Hyperspectral Imaging Widely Implemented and How to Change That?, Netguru