A few weeks back, I decided to participate in a Hackathon with a few friends. The topic was based on developing tools to make farming more efficient and sustainable.
The event was unfortunately canceled 🙁, but the topic continued to intrigue me.
I found myself diving more into the topic and would love to share my learnings.
The problem statement we came up with was: 'Using drones with multispectral & thermal imaging, geospatial data, and AI to transform the current blanket farming into a sustainable, precision-based agriculture system.'
Current farming practices rely on 'blanket applications' of fertilizers, water, etc. However, this is not a sustainable practice. Imagine the impact excess use of fertilizers could have on the quality of soil in the long run. Think of the amount of water that is wasted on irrigating without direction.
Image from Google
Now coming to the interesting metric or solution is based on. NDVI stands for Normalized Difference Vegetation Index.
A score >0.33 indicates rich vegetation + high chlorophyll content.
A score <0.33 lies in the red-brown spectrum indicating poor vegetation.
My Architecture diagram
The architecture for our solution is simple -
🌿 Drones equipped with multispectral and thermal cameras capture detailed images of the field.
🌿 These images will then be processed into NDVI maps and thermal maps (using tools like QGIS).
🌿 The maps are combined using geospatial tools (like GeoPandas) to show variations across the field by tagging them location-wise.
🌿 AI/ML models analyze the data to identify crop stress, soil issues, and pest risks by classifying it.
To discuss the kind of tagging/labelling for AI models -
Ex: A low NDVI score + high temp ( captured using thermal map data) —> low moisture area —> hence need more irrigation.
🌿 The system provides real-time, zone-specific advice for targeted spraying of fertilizers, or irrigation.
🌿 Farmers receive valuable insights via a dashboard, enabling smarter, affordable precision farming.
The solution is modular and accessible, especially with the current advancements in drone technology.
AI generated visual
For now, I wanted to share all of this, I am keenly looking into datasets, etc which will help me build a functional model.
I am looking forward to any insights from you 😊
Top comments (0)