A novel research pipeline designed to detect river and water body subtle modifications
EU SPACE TECHNOLOGIES: Using MUNDI platform
SATTELITE DATA FROM: Sentinel-2, Sentinel-3, Sentinel-5P (Did not manage yet to obtain data from Sentinel-5P)
A pipeline was designed, yet not entirely implemented due to its complexity and multitude of non-related tools used to attempt to detect and alert subtle river and water body modifications that could potentially lead to unnatural changes in the nearby geographic topology and immediate network of connected rivers.
Starting from the NetCDF Data format of Sentinel-2 and Sentinel-3, we feed it into a Mask R-CNN (matterport implementation) NN using a Convex Hull Algorithm for building up the mask on rivers in an area of interes (Bistroe Channel). We train it on manually annottated data if no better option is found, with the hope of allowing the CNN to detect river outlines. We further look to find an *average* of the shapes at different time points and feed the data into CRIMSON (used for blood vessel research)...