Image detail: The three phases of deep learning that illustrate how a deep network learns features through training and validation before being tested on new data.

Traditional techniques for classifying satellite pixels largely rely on spectral, textural measures, and sometimes temporal measures. However with Petabytes of satellite data that is now available machine learning techniques are becoming useful. From Artificial Neural Networks to Deep Learning techniques the earth science community is exploring ways to reduce time towards feature detection and increasing accuracy. We are using a variety of methods to classify dust, smoke and other features in satellite imagery.

Reference : Pullman, M., M. Maskey, R. Ramachandran, S.A. Christopher, I. Gurung, 2019: Applying Deep Learning to Hail Detection: A Case Study, IEEE Trans. on Geoscience and Remote Sensing, 10, 57(12), 10218-10225