What is Hyperspectral Imaging?
Traditional photography samples the electromagnetic spectrum at three points (bands) in the visible range, red, green and blue (RGB). Hyperspectral cameras are capable of sampling the electromagnetic spectrum in bands far outside of the visible range. For a camera to be a hyperspectral camera, as opposed to a multispectral camera, the instrument must be capable of sampling hundreds of bands.
All cameras produce images which can be thought of as three dimensional, where the 3rd dimension is wavelength. For traditional RGB cameras the 3rd dimension will only represent 3 bands, but for hyperspectral images the 3rd dimension will represent hundred of bands of near-continuous sampling. For this reason, hyperspectral images are best though of as data cubes.
Geological Survey Core Scanning Suite
Geological Survey has a core scanning suite consisting of a Short-Wave Infra-red (SWIR) camera and a Medium-Wave Infra-red (MWIR) camera, along with a high resolution RGB camera.
The SWIR camera has a spectral range of 978-2580nm sampled across 288 bands. The MWIR camera has a spectral range of 2712-5274nm samples across 154 bands.
We have over 400km of drill core in our core store and are in the process of scanning all of it. We currently have SWIR and RGB data for over 1000 holes, and MWIR data for over 500 holes. Below is an example of the data available from our SWIR camera.
A sample dataset consisting of 1 box of core is available from the
downloads page of this website.
The entire dataset is freely available please contact gsi.corestore[AT]gsi.ie to discuss.
Processing Hyperspectral Images
A single box of core is represented by ~1.5Gb of data once scanned by all three cameras. Hyperspectral images are in the "Big Data" domain, which brings it's own challenges to working with them.
Before the images can be interpreted a number of corrections and pre-processing steps must be completed. There are a number of commercial software packages designed for this, and there are also a number of open source python packages. For a brief introduction to pre-processing, Geological Survey has published notebooks in our github repository which use the downloadable sample data as an example.
Applications of Hyperspectral Image Analysis
The Geological Mapping Programme is working with the hyperspectral data to demonstrate the potential of hyperspectral core scanning to improve stratigraphy, mineral exploration, geothermal exploration and potentially improve our geological map products.
We still have a projects underway to explore the aplications of hyperspectral image analysis to understand stratigraphy, but one successful project developed hyperspectral facies that can be applied to the Irish Carboniferous. More details of the project can be found in
this paper.
We are studying the applications of hypespectral image data in combination with downhole geophysics and other techniques to geothermal exploration. Some preliminary results were release in
this poster.
We are also funding cutting-edge, multi-disciplinary research by partnering with UCD's Spectral Imaging Research Group, leveraging their data analysis and hyperspectral expertise for geoscience applications