technical writing

1

HOW A SPECTROGRAPH GATHERS & PROCESSES DATA

Spectral imaging provides access to a world beyond the scope of the human eye. With this

specialized technology, we can identify and distinguish landscape features, collect data on

distant stars, or even read the pages of an ancient book without opening the cover. Sometimes

referred to as hyperspectral imaging, spectral imaging is a form of photography that “provides

a digital image with far more spectral (color) information for each pixel than traditional color

cameras”[1] . This guide will help readers understand how experts use a spectrograph to

compile and analyze data by examining three crucial components: wavelengths and

electromagnetic spectrum, data cubes, and spectral analysis.

Wavelengths & the Electromagnetic Spectrum

Spectral imaging is based on spectroscopy, the phenomenon in which white light separates into

colored bands representing different electromagnetic energies. Because light functions as a

wave, the color bands signify waves of varying length, frequency, and amplitude (height). For

example, higher energy waves have shorter wavelengths, while lower energy waves have

longer wavelengths [2, Fig. 1].

Figure 1: Spectrum of Electromagnetic Energies

2

Using these different energies, scientists can take measurements or make opaque materials

more transparent to reveal details not visible to the naked eye. To accomplish these tasks, they

use spectral imaging devices [3, Fig. 2], which are more complex versions of the original

refracting instrument Sir Isaac Newton invented over four centuries ago [4, Fig. 3].

Figure 2: Modern Spectrograph Figure 3: Newton’s Spectroscope

In Newton’s original model, sunlight enters through the window and passes through a prism,

which reflects the color spectrum onto a screen. Though the technology has advanced, the

basic process is essentially the same. The spectroscope’s lens gathers light reflected by the

target object and directs it into the spectrograph, which contains either a prism or gate to

disperse it into its component color bands / wavelengths. A sensor in the camera component

then catches the dispersed light and produces a 2D image.

Data Cubes

To develop a 3D image, multiple pictures of the target are taken and layered to create a

datacube, a series of “tens to hundreds of pictures” stacked on top of one another [1].

Researchers interpret this data by determining which elements of the target object they want

to identify, then matching them with a specific color band or wavelength on the spectrum.

They can also graph this data to track patterns or significant changes. For example, suppose a

food manufacturer wants to test the quality of their fish sticks by measuring the fat and water

content. Lab workers would focus the lens of the spectral imaging mechanism on sample fish

sticks from the required batch, and bombard them with electromagnetic radiation. The

spectrograph will disperse the reflected energy to the camera, which produces multiple

3

pictures of the fish sticks with different colored regions to represent the elements the

researchers want to assess (in this case, water and fat). The manufacturer can then use this

data to make any necessary adjustments to the composition of the product.

Spectral Analysis

With the data collected, experts must now decide how to visualize and interpret it. This is no

easy task, since spectrographs produce much larger, more complex files than a regular digital

camera that can only be analyzed with specialized software. Preferred methods may vary by

field and purpose—for example, a forensic scientist might want to identify the chemical

composition of an accelerant used in an arson case, while geologists might search for mineral

deposits by studying anomalies in spectral images of a piece of land. Some current methods for

processing spectral data include similarity mapping, unmixing, and spectral estimation.

Figure 4: Three Methods for Processing Spectral Data

Similarity Mapping In this method, scientists identify distinct features of the target, then determine the “average spectrum” for each of this features. This data is saved in a database and serves as a baseline reference for comparing/contrasting other, similar samples [5].

Unmixing This strategy allows experts to “clean up” a spectral image and collect more complex data by using an algorithm to distinguish a specific trait or signal in each pixel of the image [6].

Spectral Estimation Most useful for compiling statistical data, this approach focuses on the frequency of a particular signal and how it changes over time [5].

Depending on the method they used, scientists then feed the collected data into specialized

software such as Spectra Plus or Vernier Spectral Analysis to produce detailed graphs, maps, or

other visuals. These images allow them to identify crucial patterns in the data. For example, in

2016, a group at MIT tested a prototype for a hyperspectral imaging camera that allowed them

to view the contents of a book without even opening it. Using “tetrahertz radiation”, they

produced images of letters by targeting specific pages [7, Fig. 5]. That same year, NASA used

spectral analysis to study JR1, a Kuiper Belt object just beyond Pluto, with their New Horizons

4

spacecraft; based on the lightcurve detected by New Horizon’s sensors, they created a graph

tracking JR1’s rotational phases [8, Fig. 6].

Figure 5: Hyperspectral Image Produced Figure 6: Graphical Representation of by MIT Spectral Analysis Conducted by NASA

To create their images of individual pages and letters in the document, the MIT team used

unmixing strategies to identify the unique properties of the paper material and ink, and

highlight the specific elements they want to uncover. This type of non-invasive analysis makes it

possible to read texts that are faded, smudged, or too fragile to physically handle. NASA, on

the other hand, used similarity mapping to examine the characteristics and movements of JR1.

Their scientists compared and contrasted hundreds of images from New Horizon and the

Hubble Telescope to assess its magnitude and rotational phrases over the course of a year. By

analyzing this data, they hope to uncover clues to the origins and significance of Kuiper Belt

Objects in our solar system.

Conclusion

Like its more primitive counterpart (based on sunlight and a prism), a modern spectrograph

collects light reflected by an object, passes it through a sensor that breaks the light into its

component wavelengths, and produces a digital image that experts can use to analyze the data

and identify patterns. This process is invaluable, as it allows climatologists to track atmospheric

anomalies, criminologists to detect arson or forgery, and manufacturers to assess the quality of

materials used in their products. Recent studies have also demonstrated strong potential for

more refined use of spectral imaging in the medical field, such as conducting retinal scans,

identifying arterial plaque, and detecting cancerous growths in breast tissue [8].

5

References

[1] Resonon. (3 March 2017). “What is spectral imagining and why should I use it?” Available: https://www.resonon.com/whitepapers/Resonon-Hyperspectral-Tutorial.pdf [2] University of Arizona. (28 February 2017). “What is spectroscopy?” Available: http://loke.as.arizona.edu/~ckulesa/camp/spectroscopy_intro.html [3] T. Gilchrist and T. Hyvarinan. (3 March 2017). “Hyperspectral imaging technology:

a look at real life applications.” Gilden Photonics. Available: http://www.gildenphotonics.com/hyperspectral-imaging-/hyperspectral-imaging- technology.aspx

[4] Chem Team. (27 February 2017). “Spectrum-Newton.” Available: http://www.chemteam.info/Electrons/Spectrum-History.html

[5] Y. Garini., I. Young, and G. McNamara. (3 March 2017). “Spectral Imaging: Principles and Applications.” International Society for Analytical Cytology. Available: https://pdfs.semanticscholar.org/de63/94361a09cb9f92838c41d767933f3f0143b6.pdf [6] T. Zimmermann. (3 March 2017). “Spectral Imaging and Linear Unmixing in Light Microscopy.” Available: https://pdfs.semanticscholar.org/b388/899ba24e7db791a8794aa5e5fe8040da457f.pdf

[7] L. Hardesty. (26 February 2017). “Judging a Book Through Its Cover.” MIT News. Available: http://news.mit.edu/2016/computational-imaging-method-reads-closed-books-0909

[8] S. Porter. (27 February 2017). “New Horizons: Getting to Know a KBO.” NASA. Available: https://blogs.nasa.gov/pluto/author/wkeeter/ [9] B.I. Grammatikov. “Modern Technologies for Retinal Scanning and Imaging: An Introduction for the Biomedical Engineer.” Biomedical Engineering Online. vol. 13, no.1, pp. 1-57, April 2014.