agindex

Compute vegetation indices (NDVI, NDRE, SAVI, GNDVI, EVI, MSAVI2, OSAVI, NDWI) from multi-band drone and satellite GeoTIFFs from the command line.

Built and maintained by ag-gis — Precision Ag Python & GIS Automation.

Why

Calculating a vegetation index from a drone orthomosaic should be one command, not a notebook. agindex is a small, focused CLI for that workflow:

• Reads any GDAL-supported multi-band raster (GeoTIFF, COG, …).
• Preserves CRS, transform, and nodata in the output.
• Emits a tiled, deflate-compressed float32 GeoTIFF ready for QGIS / ArcGIS.
• Optional PNG preview with a colormap for quick visual QA.
• Sensor presets for common drone payloads (Micasense RedEdge, DJI P4 MS, Parrot Sequoia) plus explicit --*-band overrides.

Install

pip install agindex            # core CLI
pip install "agindex[viz]"     # adds matplotlib for --png previews

Requires Python 3.10+, numpy, rasterio, click.

Quick start

The repo ships with a tiny synthetic 5-band GeoTIFF you can run on immediately:

git clone https://github.com/ag-gis/agindex.git
cd agindex
pip install -e ".[viz]"

agindex compute samples/field_5band.tif \
    --index ndvi \
    --preset micasense-rededge \
    -o out/ndvi.tif \
    --png out/ndvi.png

Output:

wrote out/ndvi.tif  (NDVI, 192x128)
  count=   24576  min=-0.0432  max=+0.8910  mean=+0.5023  std=0.3072
wrote out/ndvi.png  (PNG preview, colormap=RdYlGn)

Usage

agindex list                  # show supported indices and formulas
agindex presets               # show sensor band-ordering presets
agindex compute INPUT.tif --index NDVI --out result.tif [options]

Common options

Option	Purpose
--index	One of: ndvi, ndre, gndvi, savi, osavi, evi, msavi2, ndwi
--preset	micasense-rededge, dji-p4-multispectral, parrot-sequoia, rgb
--red-band etc.	Per-band 1-indexed override (wins over the preset)
--png	Also write a PNG preview
--colormap	Matplotlib colormap for the PNG (default RdYlGn)
--stats/--no-stats	Print min/max/mean/std after writing (default on)

Examples

NDVI from a DJI P4 Multispectral orthomosaic:

agindex compute ortho.tif --index ndvi --preset dji-p4-multispectral -o ndvi.tif

NDRE with explicit bands (e.g. a custom stack):

agindex compute stack.tif --index ndre --nir-band 5 --rededge-band 4 -o ndre.tif

EVI with a quick PNG for the field report:

agindex compute ortho.tif --index evi --preset micasense-rededge \
    -o evi.tif --png evi.png --colormap viridis

Supported indices

Short	Name	Formula	Uses
ndvi	NDVI	(NIR − RED) / (NIR + RED)	General vigor
ndre	NDRE	(NIR − REDEDGE) / (NIR + REDEDGE)	Canopy N, mid-late season
gndvi	GNDVI	(NIR − GREEN) / (NIR + GREEN)	Chlorophyll at high LAI
savi	SAVI	(1+L)(NIR−RED) / (NIR+RED+L), L=0.5	Low-LAI / bare-soil scenes
osavi	OSAVI	(1+0.16)(NIR−RED) / (NIR+RED+0.16)	Robust across canopies
evi	EVI	2.5·(NIR−RED) / (NIR + 6·RED − 7.5·BLUE + 1)	Atmospheric / soil resistant
msavi2	MSAVI2	0.5·(2·NIR+1 − √((2·NIR+1)² − 8·(NIR−RED)))	Self-adjusting SAVI
ndwi	NDWI	(GREEN − NIR) / (GREEN + NIR)	Surface water (drone form)

Run agindex list for the canonical list at runtime.

How it handles nodata

Inputs are read as float64 and pixels equal to the source nodata become NaN. Indices propagate NaN through divisions and sqrt. On write, NaNs are encoded as -9999.0 with nodata=-9999.0 in the GeoTIFF profile so downstream GIS software treats them correctly.

Use as a library

import numpy as np
import rasterio
from agindex import compute_index

with rasterio.open("ortho.tif") as src:
    red = src.read(3).astype("float64")
    nir = src.read(5).astype("float64")

ndvi = compute_index("ndvi", {"red": red, "nir": nir})

Development

git clone https://github.com/ag-gis/agindex.git
cd agindex
python -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"
pytest
python samples/generate.py   # regenerate the sample GeoTIFF

License

MIT — see LICENSE.