web analytics

LLM a powerful tool for agri analysis

In May 2024 Itsaso Guembe Cia, Henk Janssen and Eva Stierman attended the first ICE conference in Rome to present NEO’s Large Language Model (LLM) to analysis geodata. 

Use case 

Crop Classification is one of the activities involved in most of AMS analysis. However, the results are sometimes not significant enough to confirm a farmer declaration. This case is especially prevalent in parcels/ FOIs declared as trees, orchards or other woody vegetation, as they often display undergrowth (bushes, grass, herbs…) that accounts for more surface than the tree crowns themselves.

At NEO, Random Forest- based Crop Classification results are analyzed together with our National Tree Registry for the Netherlands (https://www.neo.nl/boombasis/). With the help of the NEO Assistant, users can identify the parcels with low detection probabilities and request the trees and woody vegetation that overlap them, creating a tree density figure linked to each parcel. This way, farmer declarations linked to those crops can be confirmed or declined.