Advanced Illumination for Machine Vision: Multispectral and Hyperspectral Imaging
Welcome back to our series of machine vision blogs on Illumination. In our final instalment of this series, this week we will be exploring the use of multispectral and hyperspectral imaging techniques for machine vision systems, and how illumination plays a key role in both these complex approaches to solving machine vision problems.
Whether you’re interested in building your own machine vision system for a project, a commercial application, or are just interested in learning more about illumination for machine vision, this blog post is for you.
Ultimately, illumination is fundamental for all machine vision applications, and this list of course includes printing and packaging, robotics, industrial automation, medicine, life sciences, and the automotive industry, just to name a few. We hope you enjoy reading.
Multispectral and Hyperspectral Imaging
Multispectral and hyperspectral imaging techniques have been gaining popularity over the last few years; this is because the cost of the technology has fallen, and also because new applications have arisen that benefit from these approaches.
What is Multispectral Imaging? What is Hyperspectral Imaging?
Multispectral and hyperspectral imaging techniques make it possible to image an object with multiple wavelengths.
Both approaches allow for variations in objects (that, with a naked eye, would appear very similar) to be easily distinguished and detected.
For instance, in fruit grading, it is possible to identify varying ripeness with hyperspectral imaging. This is the next level of providing early quality detection and optimal sorting and classification.
Defining Multispectral and Hyperspectral Imaging
While no strict definitions exist regarding the precise distinctions between multispectral and hyperspectral imaging, the general consensus around what separates them is the number of detection bands used in each technique.
Hyperspectral imaging generally has far more detection bands than multispectral imaging; filtering is commonly used in multispectral imaging alongside a standard RGB-IR illumination system with 4 detection bands.
On the other hand, in hyperspectral imaging, a dispersive element such as a prism or grating can be used to spread light across an array of sensors, resulting in hundreds of detection bands.
Examples of Multispectral and Hyperspectral Imaging
With multispectral imaging, we can inspect a single product in multiple wavelengths. If we take a bank note for example, this is a complex object and so to carry out full 100% inspection, we would need to inspect in at least 4 different bands: Red, Green, Blue and Infrared (IR).
Take grain sorting as a different example. If we had many different types of grain, that could well require a far greater number of different wavelengths – or spectral bands, as we also call them – so hyperspectral imaging is probably a more suitable approach.
Key Applications of Multispectral and Hyperspectral Imaging
In some applications, lots of spectral data is required to determine the exact nature of the material. This is where multispectral and hyperspectral imaging techniques can excel.
A few of the main applications where these techniques are significantly advantageous are:
- Food Sorting
- Currency Inspection
- Sorting of Recyclable Material
- Print Inspection
These plastic bottles at the recycling plant all look very similar – but their molecular compositions are different, and therefore must be separated. Hyperspectral and multispectral imaging will be able to help us distinguish between them more easily.
Left: visible image of rice infested with maggots. Right: colour applied to the same picture via image processing algorithms to easily distinguish between materials
Hyperspectral and multispectral imaging can be used to differentiate between manmade materials like polyester, and plant- or animal-based materials such as cotton or silk.
Multispectral or Hyperspectral System Checklist
All multispectral and hyperspectral imaging systems will comprise four key elements:
- Lighting source
- Target object (as described previously, will most likely be made up of varying molecular compositions but will appear uniformly similar).
- Machine vision system: either a Hyperspectral camera, or a (usually) monochrome camera and lens combined with an array of filters, or a camera with a prism and two or three sensors within the camera.
- Image processing: machine vision SDK to process material differentiation within target object(s).
Benefits of LED Illumination for Advanced Machine Vision Lighting
LEDs are the best form of illumination for practically all machine vision applications due to their low price point, energy efficiency and wide availability, making it far more practical to achieve precise broadband illumination with a combination or LED light sources of varying wavelengths.
With LEDs, it’s also possible to boost specific spectral ranges if the application calls for it. This ability is particularly advantageous in applications where the lighting system is required to match a specific spectral profile.
Considerations for LED Selection
When it comes to choosing the right LED system for a hyperspectral or multispectral application, there are four main considerations to make:
- Light Size
The physical profile or your light will impact the array density. This will have knock-on effects on the lighting output and which optics you choose. - Power
LED illumination is widely available in a vast number of different current and voltage options, as well as different polarities that apply to this wide variety of lights. You need to consider whether your environment can support the power requirements of your lighting system. - Camera & Lens
Your choice of optics may impact output efficiencies, supported spectral range, and peak wavelength characteristics. - Lifespan & Wavelength
The lifespan of your LED lights will be directly and greatly impacted by the overall strain placed on them, as well as the chosen wavelength that they will operate at, as this will differ greatly across the spectrum.
ProPhotonix COBRA MultiSpec and HyperSpec
ProPhotonix has over 25 years of expertise in designing LED illumination. Their COBRA is configurable for both multispectral and hyperspectral imaging.
It can be stacked up to 5m in increments of 100mm modules, and offers discrete control of up to 12 wavelengths from 365-1700nm.
For more information on Illumination in Multispectral and Hyperspectral Imaging, check out ProPhotonix’s whitepaper. If you have any questions on the COBRA MultiSpec or HyperSpec, on your specific application, or for anything else machine vision related, please feel free to contact us for more information.
Choosing a Camera for Multispectral or Hyperspectral Imaging
Both multispectral and hyperspectral imaging techniques can work in visible and non-visible spectral bands. The key part now is choosing a camera that will operate within these ranges.
Multispectral line lights will work with, for example, four-channel line scan colour cameras (R,G,B & IR), or a camera with three CMOS sensors, separating Red, Green and Blue per CMOS sensor.
The XCM3C4080T3 camera from NED (Nippon Electro-Sensory Devices) is an excellent option for a three CMOS line scan colour camera.
On the other hand, the MV1-D2048x1088-HS05-96-G2 from PhotonFocus is designed for demanding hyperspectral imaging applications.
MV1-D2048x1088-HS05-96-G2 from PhotonFocus
Built with a specialist IMEC line scan wedge CMV2K-LS150-VNIR sensor, this camera can be used to detect 150 pass bands within the spectral range of 470nm-900nm. Its global shutter also makes this camera optimal for high line speed applications. And with a GenICam-compliant GigE interface, it’ll fit right into many existing imaging systems.
Machine Vision Illumination & Lighting Solutions from ClearView Imaging
That’s it for our series of blog posts on illumination and lighting techniques for industrial applications. Be sure to check out our great range of machine vision lights and filters over in our products section!
Here at ClearView, we have a broad range of knowledge and machine vision expertise to help you decide on the right solution for your project.
We offer a huge range of components, and our experts are happy to help no matter what your question or problem may be. Feel free to get in touch with us and one of our machine vision experts will be ready to help you get going with your project!
