Hyperspectral Imaging: Understanding its Purpose and Benefits

Hyperspectral Imaging: Understanding its Purpose and Benefits

What is hyperspectral imaging?

Hyperspectral imaging offers more precise colour and material identification by delivering significantly more detailed information for every pixel, surpassing the capabilities of conventional imaging methods such as a colour camera. Unlike a colour camera that utilizes only three channels, hyperspectral imaging divides the light signal into numerous bands or channels, ranging from tens to hundreds. As explained below, this increased resolution greatly enhances the accuracy of machine vision, often in a remarkable manner.

Although hyperspectral imaging may sound like a novel concept, it is essentially an extension of conventional spectroscopy. Spectrometers disperse a light beam into a continuous range of "colours," commonly accomplished using a prism, for instance. The collection of colour bands forms a spectrum of the light beam, and the scientific study or practical application of light spectra is known as spectroscopy. A hyperspectral imager functions as if it were multiple spectrometers operating simultaneously, generating a spectral curve for each individual pixel within a scene, as depicted schematically in Figure 1.

hyperspectral imager light dispersing

Figure 1, a hyperspectral imager disperses the light emitted by each pixel in an image, creating a continuous spectrum that offers comprehensive data necessary for precise object identification or classification within the scene.

Why is it beneficial?

Unlike the human brain that relies on only three primary colours perceived by the human eye, computer vision systems can leverage a significantly larger number of colour channels. For instance, let's examine the colour image of two varieties of candies illustrated in Figure 2. One type of candy is arranged in the form of an "I." Conventional colour imaging systems would encounter substantial challenges in distinguishing between these two candy types, as they possess similar colours, much like many humans struggle to differentiate them.

candy type colour image In Figure 2, we observe a colour image featuring two distinct candy types, with one type arranged in the centre to form an "I" shape.

Figure 3 displays the spectral curves corresponding to the two candy types. It is evident that there is a significant amount of overlap in the curves, as expected when dealing with objects of similar colours. However, there is also a noticeable region where a distinct difference can be observed.

spectral curves-candy typeIn Figure 3, the spectral curves of the two candy types depicted in Figure 2 are presented. The average curves are represented by the dark blue and green lines, while the shaded regions indicate the standard deviation. It can be observed that the curves largely overlap across the spectral range, but they diverge from each other between 550 nm and 600 nm in terms of wavelength.

Machine learning algorithms easily leverage the distinctions present in the spectral curves illustrated in Figure 3 to precisely classify the candy types. A false colour classification map is provided, where the false colours correspond to the colours of the curves displayed in Figure 3.

coloured candies-hyperspectral dataFigure 4 displays the classification map for the coloured candies featured in Figure 2. Using hyperspectral data, the candy types are accurately distinguished through a pixel-by-pixel classification process. It is worth noting that nearly every non-glare pixel is correctly classified.

In conclusion, hyperspectral imaging offers enhanced information per pixel, making it particularly valuable in discerning between objects or materials that share similar colours. The outputs can be seamlessly integrated with robots, air-jets, labelling devices, and more. Similar to the capabilities of the human eye, hyperspectral imaging finds application in a broad range of fields, including quality control (lumber, textiles, paper, building materials, drugs), process control (thin films, moisture content, colour), sorting (food, recyclable materials, minerals), remote sensing (ocean colour, environmental monitoring, agriculture), and various others.

With the advancements in compact, cost-effective, and robust hyperspectral imaging systems, this technology can be utilized in diverse environments and across platforms ranging from microscopes to airplanes.

Hyperspectral Cameras

Resonon's hyperspectral cameras are advanced imaging devices that operate on a line-scan principle, also known as "pushbroom" imaging. These cameras are designed to scan a wide range of wavelengths, including the visible, ultraviolet, and near-infrared spectra. They possess remarkable durability and cost-effectiveness while delivering superior precision, minimal distortion, low stray light, and exceptional image quality.

 

Pika UV

330 - 800 nm
Ultraviolet + Visible

Pika L

400 – 1000 nm
Lightweight, Compact VNIR

Pika XC2

400 – 1000 nm
High-Precision VNIR

Pika IR

900 - 1700 nm
High-Speed Infrared

Pika IR+

900 – 1700 nm
High-Precision Infrared

Pika IR-L

925 – 1700 nm
Lightweight Infrared

Pika IR-L+

925 – 1700 nm
Lightweight, High-Precision Infrared

These Resonon hyperspectral cameras are versatile and can be seamlessly integrated into various Hyperspectral Imaging Systems offered by Resonon. These systems cater to a wide array of applications, including laboratory research, outdoor environments, remote sensing, machine vision, and optical sorting tasks.

When you opt for Resonon's research systems, they come equipped with Spectronon software, which provides comprehensive functionality. Additionally, for machine vision applications, we also offer real-time deep learning software. To facilitate the integration of Resonon hyperspectral cameras into different software environments, Resonon provides a programming guidance document that supports readily-available software development kits (SDKs) in multiple programming languages.

For further information on how hyperspectral imaging can address your specific application, please contact us.

Contact us to find out more on our products:

    Name*:

    Company*:

    Email*:

    Mobile No*:

    Country*:

    Product Enquiry: