Home / Projects / Crop Asset Intelligence
Decision tools · In progress

Crop Asset Intelligence

A satellite system that verifies what is planted in every field and reads its growth stage at scale, treating crops as assets. In development for an agribusiness producer, with crop-health scoring and yield forecasting being built on top.

Domain
Agribusiness
Client
Confidential
Data
Sentinel-2 satellite
Status
In progress
Challenge

The problem to solve

A producer manages crops across many fields and vast areas, yet checking what is planted, what stage each field is in and how it is doing means sending people to the ground, slowly and at high cost. Decisions about the producer's most valuable assets run on partial, late information.

Working system

What we are building

The system reads satellite imagery for every field and, across large areas in seconds, identifies the crop and its growth stage. Each decision is checked by agronomic safety rules rather than left to a black box, so it is traceable, not a guess. That proven core is now being extended to score crop health and forecast the productivity of each field, turning the crop portfolio into something the producer can monitor continuously.

Architecture

From orbit to decision

Satellite imagery is turned into vegetation signals, classified by a model ensemble, checked by safety rules, and resolved into the crop and its stage for every field.

Sentinel-2 L2A 12 bands · 5-day revisit NDVI · EVI · Pheno indices + MCD12Q2 Crop · Stage per-field output CNN ResNet50 · XGBoost ensemble · rule-based safety gates
Pipeline: Sentinel-2 → vegetation signal → ensemble classifier + safety checks → crop & stage.
The system

What it does

01
See
Read Sentinel-2 imagery for each field, refreshed every few days across 12 spectral bands.
02
Read
Compute vegetation indices such as NDVI and EVI that signal crop vigour and health.
03
Identify
An ensemble of specialist models classifies the crop and its growth stage, every decision checked by rule-based safety gates.
04
Scale
Cover thousands of square kilometres in seconds, turning field visits into a continuous overhead read.
Intelligence

Stack and methods

Sentinel-2 imagery Vegetation indices (NDVI · EVI) CNN (ResNet50) XGBoost ensemble Phenology (NASA MCD12Q2) Rule-based safety gates
Goal

What it will deliver

When complete, the producer gets a continuous, objective read on every field: what is planted, what stage it is in, how healthy it is, and what it is likely to yield, early enough to act. Crops become a managed portfolio rather than a once-a-season surprise.

Evidence

Implementation in progress

BUILDING

The satellite pipeline, model ensemble and safety gates are being built. Architecture docs and technical specs are available for review. A live walkthrough follows completion.

Next

How it can evolve

Build crop-health scoring and per-field yield forecasting on top of the classification core, push automated alerts on health and yield risk, and connect the read to planning, logistics and commercial decisions.

Where systems stop, Nex begins.

Too complex for spreadsheets. Too specific for generic software.

Start a project →