Project Abstract

The motivation for this project was to create an efficient way to extract water bodies from satellites and to reduce human error. Some of the applications of the project could be finding water for wildfires and for flood prevention. The aim of the project was to achieve an 85% accuracy, this accuracy goal is to set a target for the project. The Project implements image segmentation using a u-net to extract the water bodies. Some of the key finding from this project are that providing enough accuracy for more demanding applications is not possible with the u-net I implemented. Testing the accuracy of the results from the extraction against masks is an issue if the mask is also not accurate. To conclude during the project, I found that creating a water body extraction tool with sufficient accuracy would require more research into new method for image segmentation and as close to 100% accurate masks to test against.

Keywords: Computer vision, Image segmentation, Automatic water body extraction

 Conference Details

Session: Poster Session A at Poster Stand 48

Location: Sir Stanley Clarke Auditorium at Tuesday 7th 13:30 – 17:00

Markers: Joe MacInnes, Arnold Beckmann

Course: BSc Computer Science, 3rd Year

Future Plans: I’m looking for work