Using Computer Vision to Detect River Pollution

A system capable of detecting pollutants, sediment flow, and overall health of river systems by analyzing publicly available satellite imagery. 

Screen Shot 2018-01-31 at 12.00.39 PM.png

Below are a few examples how the visual data will be used for detecting pollution, algae blooms, and sediment content. 

Screen Shot 2018-01-31 at 2.18.15 PM.png


Build a computer vision system capable of tracing the path of a river as it winds downstream, and take cross-sectional image data for analysis. 

Screen Shot 2018-01-31 at 12.00.58 PM.png

On the question of the width of the cross section, it'd be cool to have maximum granularity to the pixel level, because data can always be grouped upward, but cannot be easily derived downward in granularity. 

Visual analysis of colors could be done manually, with a human-powered tool for annotating color changes. This could also be automated by using the color data (primarily the "hue" vector of HSL colorspace) to detect inflection points and sharp changes in values. 

Screen Shot 2018-01-31 at 2.18.41 PM.png

Feel free to hit up the comment section below with links to your projects or other projects that might help us build this!