What an RGB Vegetation Index Measures
Healthy leaves absorb most red and blue light for photosynthesis and reflect relatively more green — that is why plants look green to us. An RGB vegetation index exploits exactly this: it compares the green channel of each orthomosaic pixel against the red and blue channels. Where green clearly dominates, the pixel is living vegetation; where red and blue catch up, it is soil, straw, dead or yellowing plant material.
The crucial difference from NDVI is the missing near-infrared band. A healthy canopy reflects NIR very strongly, and that strong signal is what gives NDVI and NDRE their sensitivity to biomass, leaf structure and chlorophyll content. An RGB camera only sees the visible part of the spectrum, where the contrast between healthy and slightly stressed leaves is much smaller. RGB indices therefore measure greenness and green cover very well, and physiological vigour only weakly.
Most RGB indices are normalised ratios, so a uniform change in brightness partly cancels out. Partly is the key word: exposure, white balance, sun angle and shadows still shift the values more than they would with a calibrated multispectral sensor. If you want the background on the NIR-based indices first, read our vegetation indices guide and the NDVI vs NDRE comparison.
The Four RGB Indices: VARI, ExG, VDVI and NGRDI
DroneField computes four RGB vegetation indices from the orthomosaic of any supported survey drone. They all use the same three channels but weight them differently, so each one is tuned to a slightly different question.
VARI — Visible Atmospherically Resistant Index
VARI = (Green − Red) / (Green + Red − Blue). Subtracting blue in the denominator makes VARI less sensitive to haze and atmospheric scattering than a simple green–red ratio. It is the most general-purpose RGB index: a good first look at greenness and green-up across the whole field, and the closest thing to an "NDVI-like" overview an RGB camera can give.
ExG — Excess Green
ExG = 2 × Green − Red − Blue, usually calculated on normalised channels. ExG is built to separate plants from background. Green vegetation stands out sharply, while soil, straw and residue fall to low values regardless of their exact colour. That makes it the index for emergence and stand-gap checks, for green weed patches on bare or stubble ground, and for any job where the question is simply "plant or no plant?"
VDVI — Visible-band Difference Vegetation Index
VDVI = (2 × Green − Red − Blue) / (2 × Green + Red + Blue). VDVI is essentially a normalised form of ExG. Its values stay in a fixed range and it grades greenness more smoothly, so it works well for mapping differences in green cover within a growing crop rather than just plant versus soil.
NGRDI — Normalised Green-Red Difference Index
NGRDI = (Green − Red) / (Green + Red). Because it contrasts only green against red, NGRDI responds strongly when leaves turn yellow or brown. It is the RGB index for yellowing, ripening and desiccation: it separates still-green plants from dried-down ones more clearly than an index designed for plant-versus-soil.
What RGB Indices Can Do — and What They Cannot
Being honest about the limits is what makes RGB indices useful rather than misleading.
Where RGB indices work well
Emergence and stand assessment: early in the season, when soil is visible between plants, ExG and VDVI separate crop from soil very cleanly — often as clearly as NDVI does at that stage. Weed patches: green weeds on bare soil, stubble or after harvest show up sharply in ExG, which makes RGB flights a practical basis for patch spraying. Desiccation and harvest-aid decisions: NGRDI and VARI show which zones of a sunflower, rapeseed or potato field are still green and which have already dried down. Green cover and canopy gaps: hail, waterlogging, wildlife or machinery damage that removes green cover is obvious in any RGB index.
Where NDVI and NDRE are better
Vigour and biomass differences inside a dense, closed canopy: once the soil is covered, every RGB index reads "green" almost everywhere, while NDVI — and especially NDRE — still separates stronger from weaker zones. Nitrogen and chlorophyll status: the red edge and NIR bands react to chlorophyll and leaf structure long before the colour change is visible; RGB indices only see deficiency once leaves visibly pale. Comparisons across dates and fields: multispectral reflectance is corrected with the sunlight sensor, whereas RGB values depend on exposure and light, so an RGB map from a sunny morning and one from an overcast afternoon are not strictly comparable.
The practical rule: if the decision is about where green plants are, an RGB index is often enough. If the decision is about how well a closed crop is growing — for example variable-rate nitrogen in mid-season wheat or corn — use NDVI or NDRE from a multispectral drone.
Which Drones Can Produce RGB Vegetation Indices
RGB indices are calculated from the orthomosaic, so every survey drone DroneField supports can produce them — no multispectral camera needed. That includes the DJI Mavic 3 Enterprise, the DJI Phantom 4 RTK, the DJI Matrice 4E and Matrice 4T (mapped with the wide RGB camera), and the M300 / M350 RTK with the Zenmuse P1 (beta). On the multispectral platforms — the Mavic 3 Multispectral and the Phantom 4 Multispectral — you get both: the NIR-based indices from the multispectral bands and the RGB indices from the colour camera, from the same flight.
DroneField works with DJI drones only. The current list, including beta platforms, is kept in the supported survey drones documentation.
For many spray-drone operators this is the real advantage: the RGB survey drone they already own for boundary mapping and obstacle checks also delivers a usable vegetation layer for emergence checks, weed patches and desiccation — before any investment in a multispectral platform.
Flight Settings for Reliable RGB Index Maps
Because RGB indices depend on the colour the camera records, flight conditions matter more than with a sunlight-sensor-corrected multispectral survey.
Overlap, altitude and resolution
Use 75–80 % front and 65–75 % side overlap with the gimbal pointing straight down (nadir). For RGB mapping, 80–120 m above ground is the usual range; low flights at 30–60 m also stitch accurately and are worth it when you need to see individual seedlings or small weed patches. DroneField processes at a target resolution of 2–20 cm/px (5 cm by default); the finest possible value follows the native resolution of your images. Our flight settings guide and the flight planning tips go into more detail.
Light
Overcast, even light is preferred. Hard shadows darken pixels and push RGB index values down; patches of cloud moving across the field during the flight create bright and dark bands that look like crop differences. DroneField's illumination balancing evens out brightness between images, but it cannot recover detail lost in a deep shadow. If the light changes sharply mid-flight, it is usually better to pause and fly the rest under the same conditions.
Consistency between flights
If you want to compare dates, fly at a similar time of day, at the same altitude, under similar light. DroneField shows every index on a fixed colour scale, so maps from different projects and dates look comparable — but with RGB data, remember that part of any difference may be lighting rather than crop.
From RGB Index to Application Map in DroneField
The workflow is the same as for multispectral data: fly, import the flight folder, process with one button (alignment, bundle adjustment, DSM and orthomosaic), then choose an RGB index in the analysis panel. Presets for a healthy stand, stress, and water/soil set a sensible starting colour scale, and a built-in value guide helps you read the map. Processing runs locally on your computer — no upload and no per-hectare fees. The vegetation index documentation describes each index inside the app.
From any index, one click creates a gridded application map. The grid cell size (2–10 m, 8 m by default) is matched to the spray swath and can be rotated to the field's long edge. You then choose how the cells are zoned:
Mean index — for fertiliser and foliar feeding
Each cell is classified by its average index value. Lower-greenness cells can receive a different rate from greener ones. With RGB data this is most meaningful while the canopy is still open; in a closed crop the zones become flat, which is exactly the limit described above.
Living share — for desiccation and weed patches
Each cell is classified by the share of live plants it contains. DroneField computes the live/dead threshold automatically, so in a desiccation job the still-green cells get the product and fully dried-down cells can get zero; in a weed job on bare ground the cells with green plants get the herbicide.
Zone doses start at zero — you enter the rate for each zone, in l/ha or kg/ha, with up to 10 zones. Totals per zone and per field, an edge buffer that lets the drone cover the field edge, a zone buffer and a minimum valid share rule complete the plan. The finished map exports to DJI Agras, XAG and ABZ Innovation spray drones, Hardi sprayers, ISOXML terminals or Shapefile. See what variable-rate application is, the application map documentation, or how the AI-assisted planning with Luna can suggest the index, the zoning basis and the doses for you.
Reading RGB Index Maps Without Fooling Yourself
An RGB index reacts to colour, and not every colour change in a field is crop health. A few situations catch out even experienced operators.
Flowering crops: yellow sunflower heads or rapeseed flowers lower NGRDI and VARI even where the plants are perfectly healthy. Soil and residue: dark, wet soil and light, dry soil give different background values, and green-tinged residue or moss can lift ExG slightly. Shadows: tree lines, power-line poles and buildings cast shadows that read as low index values. Weeds inside the crop: once weeds and crop are both green and overlapping, an RGB index cannot tell them apart — it maps green cover, not species. That is why DroneField does not claim automatic weed identification; the map shows where green is, and the operator decides what it is.
The habit that prevents mistakes is the same as with NDVI: treat the index map as a guide to where to look, then check a few zones on the ground before committing a prescription. Overlay the index on the orthomosaic, compare it with what you saw while flying, and only then set the doses. Used this way, RGB vegetation indices turn the survey drone you already have into a genuinely useful tool for emergence checks, patch spraying and desiccation planning.