High-Level Project Summary
Our project serves to implement standard and simple methods from spatial methods of signal processing applied to InSAR technology to compare satellite data where abrupt changes in time happen. This means that the usual spatio-temporal methods that function extremely well to clean data are not suitable for our uses. Our final objective is to represent how certain points on our region move during, making it possible to visualize rapidly changing terrains, in situations like earthquakes, volcanic eruptions and grave geographical faults. While the arguments aren't that sophisticated or cutting-edge, it provides an important (didatic) step into understanding these methods.
Link to Final Project
Link to Project "Demo"
Detailed Project Description
At a rough level, our project takes in data SAR, puts it into a format we can manipulate, disregards highly incoherent sections of data, and represents the interferograms so we can analyze changes through time. We have implemented it in Jupyter Notebook, an opensource notebook, with an Ipython kernel. We used CUDA to make everything fast.
Space Agency Data
https://search.asf.alaska.edu/#/?dataset=SENTINEL-1%20INTERFEROGRAM%20(BETA) (from NASA)
Hackathon Journey
We were a very inexperienced team with signal processing, (In)SAR data and almost all of the technical aspects of this challenge, but we wanted to truly challenge ourselves and get out of our comfort zone. It is hard to quantify just how much we learned about signal processing, about convolutions, about project management, about python, about notebooks and everything related to that.
We were stuck for more than 4 hours on just loading the data, and a few more to realize that we needed the coherence, phase and amplitude and a few more to understand how to incorporate all the information we had.
References
Almost all of the algorithms and analysis was taken from the ESA book on the subject provided by the challenge. https://earth.esa.int/documents/10174/2700124/sar_land_apps_1_theory.pdf