If You Can Distinguish a Mineral With Your Eyes, AI Can Find it Too
We've been asked so many times, "How do your mineral maps DO that? The ‘insert your favorite mineral here….Albite, Coal, Pyrite, Calcopyrite, Cubanite, other-ites….’ doesn't have an absorption feature in the hyperspectral range you are scanning!"
And that is true. While those minerals don't have obvious hyperspectral absorption features, they do have signals, and that is good enough for us.
Example of Ugly SWIR data - Pyrite in Oilsands
This is just one of the reasons we’ve created so much technology that perfectly pairs imagery to hyperspectral sensors. We understand that geologists want, and need, everything mapped in a way that makes sense visually. After all, real eyes are good (seeing is believing) !!
So, it is time to change how we think about this. The old methods of hand processing hyperspectral absorption features are outdated and lead to noisy mineral maps. The original pioneers of this field used these tools because at the time, they were cutting-edge. However, today we have rendered hand processing of data obsolete because of the significant advances in Artificial Intelligence (AI) and the amazing tools and innovations that continue to be advanced by tech giants, such as TensorFlow (a free, open-source software library for dataflow and differentiable programming across a broad range of tasks).
Below is a video showing how we can hand-train AI. In this example, we identify biotite in core, and let our AI do the rest. This allows us to transcend from only a few pixels of hyperspectral information to millions of pixels in a library relatively quickly.
Example: Using Signal Processing to find Biotite
Biotite is easy to distinguish, but other minerals may not be. At Enersoft, we work closely with our customers to make and present fit-for-purpose mineral maps. Here’s a fun challenge: see if you can spot the Amphibole in the image below. Once you see it appear, test yourself to see if you can find it again.
Find the Amphibole
Another funny question we frequently get asked: "What software are you using to do that ?" Our answer is, "A million lines of python, or maybe Microsoft Paint".
So, the real question is: “If it can’t be seen with the naked eye, and it has no discernable hyperspectral features, does it actually exist?” Kind of reminds us of the old adage, “If a tree falls in the forest, does it even make a sound?”
Tune in next time for more great learnings and subscribe for updates.