Addressing Drilling Earth's Past Project
The automatic detection of veins was approached by transforming the cubic .img files into a series of JPEG images that would then be analyzed using OpenCV. OpenCV's process was written such that each image was slightly blurred, greyscaled, a filter was added for threshold values where the pixel differences would be more extreme. After this we added a Principal Component Analysis (PCA) stat procedure to identify "semi-linear" "lines" and then drew contours around those.
Classifying the Mineral Spectra was solved by removing wavelengths prone to distortion from artifacts from the .img files (wavelengths less than 1000, more than 2400) and then normalizing and smoothing the specrtum data. We collected 20 random samples from the .img files by asking our stakeholder Rebecca to label them, which gave us a clean, labelled data set we could use to validate our unsupervised model. With this dataset, a model was built out to then compare against the .img wave spectrum files and similarity score rankings were returned for each mineral in our mineral spectrum dataset.
Install:
pythonversion>= 2.7.13apt-get install python-pipapt-get install python-cvapt-get install ffmpegapt-get install zlib1g-devapt-get install python-tkapt-get install libsm6 libxext6
pip install numpypip install spectral