Ecosystems around the world are under immense pressure from assorted human activities. As the world’s population rises, there is an ever increasing need to feed larger and larger populations, which often comes at an expense of forested areas. Deforestation leads to a slew of other problems, including increasing CO2 levels, soil erosion, ecosystem collapse, etc.
Using the latest satellite imaging and GPS technology in tandem with image recognition algorithms we can detect deforestation at scale in a timely fashion and create alerts that can be handed over to the appropriate authorities.
While satellite images on Google Maps/Earth are often a year or two out of date, we can use these for the purpose of the event. A clearly defined forested area would be used as a training set, where we can manually flag up forest clearings, immature forrest, mature forrest, etc. Subequently, the trained model can be applied to a larger area, where it attempt to classify what‘s on the ground and identify what potentially could be clear cutting of a mature forrest.
- Code to retrieve satellite images from Google Maps (or similar).
- Image processing code.
- Creation of a valid training set.
- A trained model.
- Summary of preliminary results on the test set.
- Abraham Yusuf
- Adaugo Okafor
- Benjamin Abegunde
- Chinonso Ani
- Michael Mgbame
- Olumide Olugbemiro