Crops you can read

Crops you can read

How Omniris turns a multispectral drone flight into calibrated crop-health maps, lets you build and save your own vegetation indices, measures every zone of a field, and shows how each one changes from flight to flight.

Summary

A multispectral camera photographs a field in several narrow bands of light at once, including light the eye cannot see, such as near-infrared (just beyond red). Healthy leaves reflect near-infrared strongly and absorb red, so comparing the bands shows how a crop is doing, often before a problem is visible in an ordinary photo. The comparison is made with a vegetation index: a short formula that turns the bands into one number per spot on the ground.

Omniris takes a multispectral flight from upload to answers. It recognises the camera's files, corrects them for the light on the day, stitches them into maps, computes the standard indices, and lets you add your own. On an agricultural site it then measures every zone you draw, flight after flight, so you can see which parts of a field are improving and which are falling behind.

Cameras and files

Upload the flight's images into a folder and start a map from it. Before anything is charged, the map dialog looks at the files and says what it found: Detected: Multispectral Files, or Detected: Multispectral + RGB Files when the drone also carried an ordinary colour (RGB) camera.

Cameras recognised

A set is treated as multispectral once at least three band files are found. If a camera names its files differently, you can still tick Multispectral yourself.

Choosing what to map

Two checkboxes decide what is made. Each is ticked automatically when that kind of imagery is found, and hidden when it is not.

Next to each box the dialog shows how many images it will use and what it costs, so you can see the price of each choice before you start. Thermal frames from the same flight never mix into these maps; they have their own option, described in the companion paper Temperatures you can measure. How a map is made, what it costs and how accurate it is are covered in Maps you can measure.

Radiometric calibration

The brightness a camera records depends on the sun as much as on the crop. A cloud passing over halfway through a flight, or a flight at noon one week and late afternoon the next, changes every pixel. Radiometric calibration removes that effect, turning raw brightness into reflectance: the share of the light reaching the ground that the surface reflects, on a scale from 0 to 1. Reflectance is what makes one flight comparable with another.

The calibration badge

When you start a multispectral map, the dialog reads the image tags and shows a Radiometric calibration card saying which correction the flight will get. Nothing needs to be set: a sunlight sensor, or a panel the camera has marked, is applied automatically when it is found.

How a panel is found

• Confirmed from the tags. MicaSense cameras mark the panel photos and record the panel's known reflectance in each band. When those marks are present the panel is certain, and it is used automatically with the panel's own values.
• Suspected from the flight pattern. When the marks are missing, Omniris looks at the first and last few captures of the flight. A capture taken well below the survey height and separated from the survey by a long pause looks like a panel shot on the ground. Because this is a guess, the card says Possible reflectance panel and offers a box, Treat this flight as panel-calibrated, to tick only if you really did shoot a panel. Nothing is assumed without your say-so.

When you select several flights at once, calibration is shown and applied per flight, one line per date. One flight with a panel never makes the others look panel-calibrated.

Calibration travels with the results

Every reflectance map made from a calibrated flight carries a label saying which calibration it had: sunlight sensor or panel. Indices you compute from those maps inherit the same label. Zonal analytics and flight comparisons also know each flight's calibration, so results from flights with weaker calibration are treated with the care they need (see Tracking change across flights).

Multispectral maps

A finished multispectral map arrives in the vault as a folder of results beside your images. Each result is a GeoTIFF (the standard map-aligned image format) and opens on the Omniris map.

Two maps per index

Every index, whether made with the map or by you later, is saved twice. The measurement map holds the exact index value at every spot, for analysis and for other software. The display map is the same thing already coloured, which is what you see on the map. Keeping both means a picture is never mistaken for the numbers behind it.

Fixed colour classes

NDVI and NDRE display maps use five fixed classes tuned for agriculture. Because the class edges never move, the same colour means the same value on every flight of the field, so two dates can be compared by eye.

The map's processing report, on screen and as a PDF, shows the NDVI and NDRE maps with this legend in the same words, so a printed report matches what your client sees in the app.

Labelled for the Site Navigator

Results are labelled automatically with the site, the customer, the flight date and the layer type (NDVI, NDRE, multispectral, red edge, near-infrared). The Site Navigator, the Omniris map for working with one site across all its flight dates, uses those labels to list each layer under its date, ready to switch on, fade and compare. The Site Navigator is described in the companion paper on Site Manager and Site Navigator.

Colour and multispectral maps are checked against each other

When a flight produces both a colour map and a multispectral map, the two are made separately and can disagree slightly about where things are. After both arrive, Omniris measures the offset between them and shows it as a small marker beside the flight date: grey when under 1.6 ft (0.5 m), amber from 1.6 to 3.3 ft (0.5 to 1 m), and red above that. Selecting it opens the detail. The offset is measured and reported; neither map is moved.

Built-in indices

Beyond the NDVI and NDRE maps made with every flight, Omniris includes eighteen standard indices. Open Band Analytics from the menu of any folder that holds a multispectral flight's band maps; the item appears only when at least two recognised bands are found. Band Analytics is available on the web and in the desktop application.

Each index is offered only when your camera recorded the bands it needs. On a four-band camera with no blue band, for example, the indices that need blue are simply not in the list, rather than shown and then failing. Values assume the bands are on the 0-to-1 reflectance scale.

The Built-ins Help link in Band Analytics opens this reference, with each index's formula, and the same reference is linked from zonal analytics. A good rule of thumb: NDVI early in the season, NDRE once the canopy fills in.

Each index states its range, and says what kind of range it is

Every built-in index carries a value range: the values it can sensibly produce. Flight comparisons leave out any spot whose value falls outside the range on either date, so one freak spot cannot swamp the result. Band Analytics says plainly what kind of range each one is:

• A mathematical limit (NDVI, GNDVI, NDRE, NGRDI, GLI, NDWI, SAVI, OSAVI): the formula cannot produce anything outside it.
• A limit set by the reflectance scale (DVI, EVI2, MSAVI2): it holds because the bands run from 0 to 1.
• A plausibility limit (SR, EVI, CI-RedEdge, CI-Green, MTCI, PSRI, VARI): these indices divide by a quantity that can approach zero, so they have no true limit. The range sits well outside published field values, and anything beyond it is treated as a blow-up. Band Analytics marks these with a warning so the range is never mistaken for mathematics.

Your own formulas

Agronomists often have an index of their own, or a variant tuned to a crop. Band Analytics lets you write it, see it on the map straight away, and keep it.

Writing a formula

The dialog lists the flight's bands. Each band has a short label (B1, B2 and so on, in order of wavelength) and, where the band is known, a name you can type instead: BLUE, GREEN, RED, REDEDGE, NIR, NIR2, THERMAL or PAN. Insert beside a band drops it into the formula for you.

A formula can be up to 500 characters and must use at least one band. It is checked as you type: unsupported characters, missing bands and unbalanced brackets are flagged before you run it. Formulas are checked against a strict list of what is allowed, so nothing in a formula can do anything except arithmetic on the bands.

Preview, then save

• CALC draws the formula live over the field, on a street, satellite or hybrid background. No result is saved until you choose to. A compute resolution slider trades detail for speed, from full resolution down to the coarsest overview.
• SAVE computes the index over the whole flight and files the result in the flight's bands folder, as a measurement map and a coloured display map. Choose a regular GeoTIFF (the default), a Cloud Optimized GeoTIFF (a GeoTIFF arranged so map software can read just the part it needs), or both. Results that were saved as regular GeoTIFFs can be converted to Cloud Optimized GeoTIFFs in place later, keeping their names, labels and links. Encrypted files and files under a legal hold or retention policy are not rewritten.
• The work runs on a processing machine sized for the flight, so you can close the dialog while it runs. Saving an index carries no processing charge.

A saved index is a full vault file: its custody history starts the moment it is created, with a record of the formula that made it, and it inherits the flight's site, customer, date and calibration labels. The display map is also labelled with the index name, so it appears in the Site Navigator under its flight date like NDVI and NDRE, even for a formula of your own.

Saved formulas

A formula you use more than once can be saved by name, with its colour scheme and, if you like, its value range. Saved formulas appear in the Band Analytics list next to the built-ins, and in the index list of zonal analytics.

• Personal or shared. Keep a formula to yourself, or share it with an organization you belong to so the whole team uses the same definition. Colour schemes work the same way.
• Who can change it. Your personal formulas are yours alone. A shared formula can be changed by the person who created it or by a manager of the organization. Built-in indices are read-only.
• Saved means trackable. Only a saved formula can be followed across flights. A one-off formula typed into the box can be previewed and saved as a map, but it has no lasting identity to compare from one date to the next.

Colour ramps and legends

An index is a number; a colour scheme decides how it looks. In Band Analytics a colour scheme is a short list of value ranges, each with its own colour: every spot from the lower value up to (but not including) the upper value gets that colour. Because the list is the scheme, it doubles as the legend: it shows exactly which colour stands for which values.

Picking a built-in index selects its scheme for you, and you can switch to any other. To make your own, add, edit or remove ranges and pick each colour, then save the scheme by name, personally or for your organization, exactly like a formula. A preview with no scheme at all is drawn in grey from −1 to 1. The NDVI and NDRE schemes are the same classes the map makes, so an NDVI you compute yourself looks exactly like the one delivered with the flight.

Legends elsewhere

• Processing report: the NDVI and NDRE figures carry the five-class legend with a plain-language meaning for each colour.
• Zonal analytics: zones are shaded on a red-to-yellow-to-green scale, with a legend giving the lowest and highest value shown.
• Flight comparisons: an index change map is drawn red where the index fell and green where it rose, centred on no change.

Reading values on the map

On an agricultural site, the numbers behind the colours are one click away in the Site Navigator, for any zone you have drawn:

• Click a zone to open its data panel. For every index computed on the site, it shows the zone's value and its spread (written as value ± spread) on each flight date you have selected, side by side.
• Click a shaded zone while zone shading is on to see its history: the value, the unevenness and the share of the zone covered by vegetation on each flight, with its trend and its standing against the rest of the field.
• Ask Orion, the Omniris assistant, where the stressed, dry or healthiest areas are; see Asking Orion.

Zone values are averages over the whole zone. To read a smaller spot, draw a smaller zone around it.

Zones and statistics

One number for a whole field hides most of what matters. Zonal analytics measures each part of the field separately: every management zone, trial strip, irrigation block or problem patch you draw. It answers two questions for each zone: how is it doing now, and how is it changing?

What you need

• An agricultural site in Site Manager, with a real name. A site still carrying the default name "Site" would mix flights from unrelated fields, so you are asked to rename it first.
• Zones drawn on the site with the polygon tools, or imported from a KML, KMZ, GeoJSON or shapefile. Draw them once; the same boundaries are used on every flight.
• Multispectral flights linked to the site, or thermal flights with a saved crop water stress layer (see Crop water stress). A site with only colour photos has no index to measure.

Open Zonal Analytics from the site in Site Manager, or from the Site Navigator. Pick an Index (only indices your camera's bands can produce are offered, including your saved formulas, plus crop water stress once it has been run on a flight of the site), a Mask and a Year, then select Compute. Any flight that does not yet have that index is computed for you first, at no charge. Later, Update new flights measures only the flights not yet done, and a new flight is measured automatically when its map finishes on a site you have already computed. Anyone who can see the site can view the results; people who can edit the site can compute and change settings. Zonal analytics works on the web and in the desktop application.

Vegetation or all pixels

Every zone holds a mix of crop, soil and shadow. The Mask decides what is measured:

• Vegetation (the default) measures only spots that are actually growing plant: those whose NDVI is at or above the vegetation threshold, 0.20 unless you change it. It answers "how healthy is the crop that is there?" The mask always uses NDVI, whatever index you analyse, so NDWI over vegetation means crop water content, not standing water.
• All pixels measures everything inside the zone. Together with cover (the share of the zone that is vegetation), it answers "how much of this zone is growing?"

Veg. threshold sets a site-wide cutoff anywhere from −1 to 1, with an optional override for any zone, and a Preview mask that paints the vegetation green and the rest grey over any flight before you save. Saving re-measures every tracked index on every flight, in the background. A thin strip just inside each zone's boundary is also left out, 6 in (0.15 m) or one pixel wide, whichever is wider (narrower only when a zone is too small to spare it), so spots that straddle the boundary and mix in the neighbouring ground are not counted.

What each zone gets

Behind these, each zone's median, 10th and 90th percentile and spread are kept for every flight. Before any of them is calculated, freak values far outside the zone's own spread are set aside, so a handful of broken spots cannot drag the average.

Alerts that find the zones to look at

• Falling behind: the zone's trend is below the field's trend. It may still look fine, but it is losing ground against its neighbours. This is usually the most useful early warning.
• Patchy: the zone is unusually uneven compared with the rest of the field, which often points to uneven emergence, irrigation gaps or a developing problem.

What you see

• A map for every flight. A row of small maps, one per flight date, with the zones shaded red to green on one shared scale and numbered. Read left to right, you can watch a zone go cold.
• A ranked table. One row per zone, with a small history chart, the latest value, trend, vs field, uniformity and alerts. Sort by any column to bring the zones that need attention to the top.
• Zone detail. Select a zone for its full history chart, its spread and cover on each flight, its season total, and its green-up and peak.
• On the real map. In the Site Navigator, Shade zones on map colours the zones over your imagery by mean, uniformity, field z-score or trend, following the flight date you select.

A Site Navigator share link shows the zones and the results already computed; the person you share it with can read them but cannot run new ones.

Management zones and boundaries

Zones are the polygons of your site, so they can be your management zones, trial plots or anything else you farm by. Their boundaries can be exported from Site Manager as a shapefile, GeoJSON, KML or KMZ for use in other farm and mapping software.

Tracking change across flights

A single flight is a snapshot. A season of flights is a story, and Omniris keeps every flight of a site lined up so you can read it.

Charts over time

• History. Every zone carries a small chart of its value across the selected year, in the table and full size in the zone's detail.
• One year at a time. Crops rise and fall every season, so one line across several years would mislead. By default you see the most recent calendar year with flights, and every trend and comparison is worked out within that year. The Year picker selects another year, or All years for the long view.
• Year over year. Tick Compare years to overlay each season's curve on a shared January-to-December axis, for the whole field or any one zone. You can see at a glance whether a zone is ahead of or behind where it was at the same point last year.

Why the comparisons can be trusted

• Calibration is respected. Absolute measures (trend, season total, green-up and peak) use only flights calibrated with a sunlight sensor or a panel. Camera-only flights are left out of them, because changes in the light would look like changes in the crop. Those flights still count for the relative measure, vs field, which compares zones on the same flight and so does not need calibration. The one exception is crop water stress from a published crop baseline: it is worked out from temperature, not reflected light, so its flights count in the trend without this calibration. The relative crop water stress is the opposite case: rescaled on every flight, it is never in the trend, even on a calibrated flight, and counts only for vs field.
• One flight per day. If a site has more than one map for the same date, the best-calibrated one is used, so a re-run cannot count a day twice.
• Complete flights only. A flight counts towards the field average only when every zone has a value for it, so a partial flight never distorts the comparison.
• The same ground every time. Your zones are reused on every flight. Edit a boundary and past flights are re-measured to match.

Comparing two flights directly

The Site Navigator's Compare tool takes any index, built-in or saved, and two flight dates, and works out the change at every spot. Any flight missing that index has it computed first.

• A change map is saved as a GeoTIFF beside the later flight, drawn red where the index fell and green where it rose.
• The numbers give the average change, the middle change, the spread, and the area that rose, fell or stayed within a change threshold (0.05 index units unless you set another). For NDVI and NDRE they also give the area where vegetation cover appeared or disappeared.
• You can compare the whole flight, a site zone, or an area you draw, and take the results away as a report or a CSV file on the web.
• If the two flights were not calibrated on a common footing (one or both camera-only, or only one with a panel), the result is clearly marked as an estimate.
• Alert me turns a comparison into an email alert: when a new flight arrives, you are told if the average change, or the share of the area that changed (10% unless you set another), passes your limit.

"Rose" and "fell" describe the number, not the crop. For PSRI (senescence) or NDWI (water), a rise can be the bad direction.

Crop water stress

A thirsty plant closes its pores and warms up. The Crop Water Stress Index (CWSI) uses a thermal map of the canopy, with the air temperature and humidity at 1 pm local time on the day of the flight, to put a number on that: 0 is well watered, 1 is maximum stress. In the Site Navigator, open a thermal layer, pick the crop and select Run CWSI.

• Crops with published baselines: corn (maize), soybean, cotton and wheat before heading. The crop must be chosen; none is assumed.
• Other crop (relative): for any other crop, a relative version where 0 and 1 are the coolest and warmest canopy on that flight, labelled as relative on the saved layer, in the Site Navigator and in the report, and kept as a separate index in zone trends and comparisons.
• Only the canopy is measured: plants are picked out with NDVI when the flight has it, at the site's own vegetation threshold (0.20 when the site has none), otherwise from the colour map.
• The result is a map in five classes, from well watered to severe, saved like any other index.
• A saved CWSI layer can be tracked zone by zone in zonal analytics and compared between two dates in Compare, even on a flight with no multispectral bands; such a flight is measured with the All pixels mask, since the layer already covers only the canopy. Unlike a band index, it is never made for you there: run CWSI on each flight first.

Thermal maps, temperatures and CWSI are covered in full in the companion paper Temperatures you can measure.

Agricultural sites

Every site in Site Manager has a Type, set when the site is created or edited. The type decides which tools the site shows.

Band Analytics, the NDVI and NDRE maps and flight comparisons work on any multispectral flight, whatever the site type. Flights join a site through the site name given when the map is made, so use the same name for every flight of the same field.

Asking Orion

Orion, the Omniris AI assistant, can answer crop questions about the multispectral flights in a folder, in plain language:

• "Show me the stressed areas", "where is there standing water", "show me the dry areas", "bare patches", "senescing areas", "the healthiest areas".
• For each flight date it picks a suitable index, marks the matching areas on a picture of the field and a heat map, and gives the statistics.
• On a calibrated flight it uses fixed, published class limits; otherwise it looks at the relative tail of the field, the bottom quarter unless you ask for another share ("the worst 10%").
• It can combine indices, for example "stressed, but not just ripening", which looks for low NDVI that is not explained by high PSRI.
• For drought and water-stress questions on thermal flights it reads the CWSI result, or works out a relative one on the spot.

The answers are worked out from your index maps, flight by flight. The companion paper on Orion, the AI assistant, describes what else it can do.

Exports for farm software

The maps Omniris makes from a multispectral flight are in open, standard formats that farm-management, agronomy and mapping (GIS) software can read.

Results can also be shared with a colleague or client, or by link, like any other vault file. Sharing, delivery and integrations are covered in the companion paper on sharing, delivery and integrations.

What the results claim, and what they don't

Crop-health maps are powerful because they are simple. That also makes it important to know what they do not say.

• An index shows where, not why. A low value says a part of the field reflects light differently. Whether the cause is nitrogen, water, disease, pests or soil is for a person to find out on the ground.
• Calibration decides what can be compared. Camera-only flights are reliable within themselves but not across flights, so they are left out of trends. A panel suspected from the flight pattern is used only if you confirm it.
• Some indices can blow up. Ratio indices (SR, EVI, CI-RedEdge, CI-Green, MTCI, PSRI, VARI) divide by a quantity that can approach zero over shadow, water or dark soil. Their ranges are plausibility limits, labelled as such, not mathematical ones.
• Your camera's bands set the menu. NDRE, CI-RedEdge, MTCI and PSRI need a red-edge band; EVI, VARI and GLI need a blue band. Indices your camera cannot support are not offered.
• Colour and multispectral maps can be slightly offset. They are made separately. The offset is measured and shown beside the flight date, but not corrected.
• No error margin for index changes. A flight comparison reports how much an index changed, not how certain that change is, because no published error model for index changes exists. Flights with mismatched calibration are marked as estimates.
• Zones are averages. Zonal results describe whole zones, not individual plants. The vegetation mask is always based on NDVI.
• Zonal analytics needs an agricultural site with a real name, zones drawn on it, and multispectral flights (or thermal flights with a saved crop water stress layer) linked to it. Seasons follow the calendar year.
• Only saved formulas are tracked. A one-off formula can be previewed and saved as a map, but is not followed from flight to flight.
• Editing needs edit access. Saving an index needs permission to write to the flight's folder, and running zonal analytics needs permission to edit the site. People with view access, including share-link recipients, see what has already been computed.
• CWSI baselines exist for four crops. Corn, soybean, cotton and wheat before heading. For any other crop the result is relative to that flight only.
• Encrypted files cannot be converted to Cloud Optimized GeoTIFFs in place, and are skipped when publishing to ArcGIS, because their stored contents are encrypted.

Companion papers

• Maps you can measure: how drone photos become orthomosaics and elevation models, and how accurate they are.
• Temperatures you can measure: thermal maps, temperature readings, hot-spot detection and crop water stress.
• Site Manager and Site Navigator: organising customers, sites and zones, and working with every flight of a site on one map.
• Orion, the AI assistant: asking questions about your sites, flights and maps in plain language.
• Evidence you can prove: the custody record, certificates, legal holds and retention behind every file.
• Sharing, delivery and integrations: getting results to clients and into other systems.
• Mission planning and flight safety: planning the flights that capture your fields.
• The phone app in the field: Omniris on a phone in the field.

Glossary

Band

One narrow slice of light that a multispectral camera records separately, such as red or near-infrared.

Cloud Optimized GeoTIFF

A GeoTIFF arranged so map software can read just the part it needs, without downloading the whole file.

Cover

The share of a zone that counts as vegetation.

CWSI (Crop Water Stress Index)

A number from 0 (well watered) to 1 (maximum stress), worked out from canopy temperature and the weather on the day of the flight.

GeoTIFF

An image file that also records exactly where on Earth it lies; the standard format for maps like these.

Multispectral camera

A camera that photographs the same scene in several bands at once, including light the eye cannot see.

NDRE

Normalized Difference Red Edge: like NDVI but using the red edge, so it keeps responding in dense canopy.

NDVI

Normalized Difference Vegetation Index: the classic crop-health index, comparing near-infrared with red. Higher means greener and denser.

Near-infrared (NIR)

Light just beyond red, invisible to the eye, that healthy leaves reflect strongly.

Radiometric calibration

Correcting images for the camera and the light on the day, so that values mean the same thing from flight to flight.

Red edge

The narrow band between red and near-infrared where a leaf's reflectance rises steeply; sensitive to chlorophyll.

Reflectance

The share of the light reaching a surface that the surface reflects, from 0 to 1. Calibration turns raw brightness into reflectance.

Reflectance panel

A target of known reflectance, photographed on the ground before or after a flight, that lets the images be calibrated to an absolute scale.

Site Navigator

The Omniris map for working with one site across all its flight dates.

Sunlight sensor

A small light meter on top of the drone that records the incoming sunlight with every photo.

Vegetation index

A short formula that combines bands into one number per spot, describing some aspect of plant health.

Vegetation threshold

The NDVI value (0.20 unless you change it) at or above which a spot counts as growing plant.

Zone

An area you draw on a site, such as a management zone, trial strip or problem patch, measured each time you fly.