Maps you can measure

How Omniris turns a drone's overlapping photos into a stitched map, elevation models and 3D, how to fly so the result is good, and every tool you can use on the finished job.

Summary

A drone mapping flight takes hundreds or thousands of overlapping photos. Stitching turns them into one accurate, map-aligned picture of the whole site, called an orthomosaic: a single image in which every pixel sits where it belongs on the map, so distances and areas measured on it are true. The same processing also measures the shape of the ground, giving elevation models, a point cloud and, if you ask, a 3D model.

Omniris does the stitching for you on processing machines sized to the job, checks the photos before you pay, files every result in the same secure vault as your photos, and gives you the tools to measure, analyse, compare and share what it made.

How stitching works

Stitching relies on overlap: each spot on the ground appears in several photos, taken from slightly different positions. From those differences the software works out where each photo was taken and the shape of the ground, much as two eyes judge depth. Omniris runs the stitching on OpenDroneMap, a widely used open-source photogrammetry engine (photogrammetry means measuring from photographs), with its own checks and steps before and after.

flowchart LR
  A[Overlapping photos
in a vault folder] --> B[Checked:
flights, angle, height]
  B --> C[Cleaned:
duplicates, tilted photos]
  C --> D[Matched and
reconstructed]
  D --> E[Surface, terrain,
orthomosaic]
  E --> F[Results folder,
report, email]
  

Very large surveys are processed in sections of up to 3,000 photo positions and then merged, and long, narrow corridor flights are split into shorter sections, so the job fits the machines that run it. Your photos are never changed.

Starting a stitch

Choose Stitch on a folder of photos in your files, or Upload & Stitch to upload and start in one step. The stitch dialog has already read the photos, so most choices arrive filled in. You confirm them, see the price, and press Start Stitching. The desktop application uses the same dialog.

Choosing a resolution

The ground sample distance (GSD) is how much ground one pixel covers. It follows directly from how high the drone flew and the camera it carried: fly twice as high and each pixel covers twice as much ground. The dialog offers the finest resolution the flight supports and will not accept a finer one, because asking for more detail than the photos hold costs far more to produce and store and returns guessed pixels, not detail. It suggests what each range is good for:

A coarser resolution is much cheaper: the size of the map grows with the square of the resolution, so halving the pixel size makes four times as many pixels. The dialog shows that consequence as you type.

What you get

Results arrive in an output folder beside the photos, named with the prefix you chose. What a job delivers depends on the camera and on the choices made in the dialog:

Every map is a Cloud Optimized GeoTIFF (COG): a standard map-image file, laid out so a viewer can read just the part on screen at the detail needed. It opens on the Omniris map and in any mapping (GIS) software. NDVI (normalized difference vegetation index) and NDRE (normalized difference red edge) are standard measures of plant health computed from how strongly leaves reflect red, red-edge and near-infrared light.

The 3D outputs of a stitch are described in the companion paper, Digital twins you can see and measure, which covers 3D models and Gaussian splats in full. Thermal maps and the temperature tools are described in Temperatures you can measure.

Flying for a good stitch

No processing can recover ground the camera never saw. A good map starts with a flight planned for it, and Omniris gives pilots the planning tool and the checks to get one.

Plan the mission

The mission planner, on the web and in the desktop application, lays out the flight lines over an area you draw. Pick the pattern for the result you want:

Forward overlap is how much each photo shares with the next one along a line; side overlap is how much neighbouring lines share. Every default can be changed. Enter either the resolution you need or the height you will fly, and the other is worked out for your camera. The planner then shows the number of photos, the resolution, the number of lines, the distance, an estimated flight time and number of batteries, and the area in acres. It warns when the speed would blur the photos, when it has had to slow the drone so the camera can keep up, and when the plan goes above the 400 ft (122 m) limit that applies to most US commercial flights.

• Keep the height above the ground constant. Over hilly ground a flight at one altitude photographs the hilltops in finer detail than the valleys. The planner tells you how much the resolution will vary across your area, and can follow the terrain, using a worldwide terrain model or a surface model from one of your own earlier stitches.
• Get the sensor orientation right. Turning the camera from landscape to portrait swaps the spacing between lines and between photos, and getting it wrong halves the overlap in one direction. The planner asks.
• Take the plan to the drone. Export for DJI Pilot 2, DJI Fly, Litchi or QGroundControl, or as a boundary for apps that plan their own route (such as Wingtra and Autel). Every export comes with a plain parameter sheet listing what to set in the flight app, including where to launch so that the heights come out right.

Check before you fly

The pre-flight checklist includes, beside the regulatory items, the checks that matter for mapping: the take-off point matches the mission plan (the plan's heights are measured from it), the storage card has room for the planned number of photos, and the aircraft has a satellite position lock before launch. The flyability forecast scores each day ahead for flying, from wind, temperature, visibility and other weather.

In the air

• Point the camera straight down for a map. This is called nadir. A tilted (oblique) photo shows the sides of things and hides the ground behind them, which a map cannot use. Photos more than 25° from straight down are set aside before stitching. A flight shot entirely at an angle cannot make a map, and the stitch dialog says so and offers to build a 3D model from it instead.
• For 3D, fly from several directions. When a flight cannot make a map, the dialog also says whether it was photographed from enough directions for a complete 3D model, or from only one.
• One site per job. If the photos in a folder come from separate places, the dialog shows each group so you can choose which to stitch. Stitching unconnected groups together is likely to fail.
• Keep positions in the photos. If very few photos carry a GPS position, the dialog warns that the stitch is likely to fail without ground control points.
• Multispectral: photograph the reflectance panel, and carry the sunlight sensor. They are what make crop-health values comparable from one flight, or one season, to the next (see "Starting a stitch").
• Ground control: place enough points, and mark each one in enough photos. The dialog counts how many photos each point is marked in. Five or more points, each marked in five or more photos, is the target it shows; fewer still works, with a warning that the fit may be marginal.

Sizing, time and price

Sized to the job

Stitching needs a great deal of memory, and how much depends on the number of photos, their size and the resolution asked for. Before a job starts, Omniris estimates what it will need and chooses a processing machine big enough for it. The resolution you choose is the main lever: a finer map needs more memory and more time, a coarser one less.

• Large surveys are split. A very large map job, or a long corridor, is processed in sections and merged, so it can run on the machines available.
• 3D is not split. A 3D model has to be built in one piece on one machine. Where processing runs on a fixed set of the site's own machines (the data center edition and on-premises), if no machine is large enough for a 3D model of the selected photos, the dialog turns the 3D options off and says why, rather than accepting a job that could never run.
• You can close the page. Once started, a job is run by Omniris, not by your browser. The job card shows the current stage and progress, the processing log can be read and downloaded at any time, and a running job can be cancelled.

How long a job takes depends on the number of photos, the resolution and whether 3D was added. Omniris does not quote a time in advance.

Price

Stitching is priced in credits, by the number of photos, and the dialog shows the price for exactly what you selected before you start.

• Failure costs nothing. A job that fails, or that you cancel, is refunded in full.
• Most tools are included. Viewing, measuring, volumes, comparing dates, terrain analysis, vegetation indices, zonal analytics, crop and sharing carry no processing charge; results they save count toward your storage. AI detection is priced separately.

Accuracy: ground control and RTK

Two kinds of accuracy matter, and they are easy to confuse:

• Relative accuracy is how true the map is to itself: whether a wall on the map is as long as the wall on site, or a stockpile's volume is right. Distances, areas and volumes measured within one survey rely on it.
• Absolute accuracy is how true the map is to the world: whether a corner on the map sits where a surveyor would put it. That depends on how well the photos' positions are known, and ordinary drone GPS is far less precise than a survey. Comparing two flights, or a flight with a design, relies on it, and the measuring tools say so: those comparisons need ground control, RTK or alignment on both surveys.

Omniris supports several ways of getting better positions.

RTK

RTK (real-time kinematic) positioning corrects the drone's GPS in flight, from a base station or a correction network, to survey-like precision. When the photos carry the drone's own record of how precise each RTK position was, Omniris tells the reconstruction, so precise positions are held tightly. RTK photos do not record which coordinate frame the corrections were in, so you can state it for the site, or for a single flight, along with whether the base sat on a known point. There is no default; the setting is recorded with the job and used when results are exported or compared with a design.

Ground control points

A ground control point (GCP) is a marker on the ground whose position has been surveyed. Mark where it appears in the photos, and the reconstruction is pinned to the surveyed positions. Omniris includes the whole workflow:

• Import the surveyed coordinates from the files that survey equipment and software produce, including Emlid Reach, Trimble Access, Leica Captivate, Pix4D, Metashape, OpenDroneMap and plain CSV, in your survey's coordinate system.
• Mark each point in the photos. Printed coded (ArUco) targets can be found in every photo automatically. A point marked once can be searched for in the nearby photos that show the same spot, and its predicted position is drawn on photos you have not marked yet, more precisely as you mark more. A mark stays exactly where you click it. Every mark keeps a history of who placed, moved or removed it, and a change can be reverted.
• Heights are put on one reference. Surveyed heights may be measured from mean sea level or from the ellipsoid used by satellite positioning, which can differ by tens of feet. Where a point's height reference is known, Omniris converts it to one common reference before the stitch; where it is not known, it warns you.
• Review the fit after the stitch: how many points and marks were used, the average error in pixels, the overall position error, and each point's error, coloured by size. Open any point to see every photo it was marked in, and re-mark a photo where the fit is poor.
• Re-stitch cheaply. After changing marks, re-stitch as a draft, a quick low-quality preview at 20% of the price, check the fit, then run the full stitch.
• Fewer than three points is flagged. With any ground control the drone's own GPS is no longer used in the solve, so a map controlled by one or two points is oriented from those points alone. The stitch dialog and the job card say so, and suggest marking more or using the points as checkpoints instead.

Checkpoints: measured accuracy

A checkpoint is a surveyed point that is deliberately left out of the fit and used only to measure it. Where the map puts a checkpoint, compared with where it was surveyed, is the map's measured accuracy at that point. The fit cannot pull a checkpoint towards its survey, which is why its error means something a control point's error does not.

• Same targets, different use. Checkpoints are surveyed, imported and marked exactly like control points. A file can be imported as control points or as checkpoints, labels that look like checkpoints are offered as checkpoints, and any surveyed point can be switched between the two; every switch is recorded in the audit trail.
• Measured on the delivered map. Each checkpoint is triangulated from its marks in the reconstruction the deliverables were made from, and the surface height under it is read from the delivered elevation model. The results give the error east, north and up for every checkpoint, the horizontal, vertical and 3D RMSE, the mean error in each direction, and which checkpoints lie outside the area the control points enclose.
• Measured again without re-stitching. Checkpoints marked after the map is delivered, or points switched afterwards, are measured on the finished map by a small separate job, with no credits and without reading any encrypted data: the stitch keeps a small plaintext record of its solved cameras and of the surface heights at every surveyed point. Every measurement is kept, with who asked for it.
• The past does not change. Every stitch records exactly which points controlled it, as the reconstruction actually loaded them. Switching a point later never re-labels or re-grades an earlier map, and a point that controlled a map is never reported as one of its checkpoints.
• RTK and GPS flights too. On a flight with no control points the map is carried into the survey's coordinate system with the same transformation design comparison uses before it is measured. When no exact transformation exists, for example an RTK flight corrected in ITRF2014 against a survey in NAD83(2011), the raw offsets are shown and the results say why.

Without surveyed points

• Ground reference points are points taken from a recognisable feature on satellite imagery, for when nothing was surveyed. They are good to about 3 ft (1 m), and a stitch that uses them is labelled accordingly.
• Align to a previous flight. A re-flight of a site can be aligned to an earlier stitch of it, so the two line up for comparison without marking anything. The result reports how closely the new surface matched the earlier one, and inherits the earlier flight's accuracy.

What the result claims

When a stitch has checkpoints, their measured accuracy is the headline: on the job card, at the top of the results, first in the stitch report's ground-control section, and in what Orion reports. Design grading uses whichever is looser, the accuracy stated for the control or the vertical accuracy the checkpoints measured, so a measurement can only tighten what is graded.

After a stitch with ground control, the results also carry an accuracy class, set by the least accurate control point used, from the accuracy stated for each point:

The stitch report (see "Sharing, exporting and evidence") states the same things in writing. It grades how many photos were placed, whether the reconstruction held together in one piece, and how closely it fits the photos. It quotes how well the result agrees with the drone's own GPS, and says beside that figure that it is not ground accuracy. When no ground control was used, it says that absolute accuracy cannot be stated from the flight alone. When a geometry check finds something wrong with a solve, it warns that elevations, contours, volumes and heights from that flight may be off, while the orthomosaic is usually unaffected.

Seeing the result

The rest of this paper is about what you can do with a finished stitch. The tools fall into five groups: see, measure, analyse, compare, and share. All of them work the same way in the web application and in the desktop application, which runs the same code; the few differences are noted where they arise.

Each kind of layer always uses the same colour scheme: terrain colours for elevation, a thermal palette for temperature, and, for change between dates, a red-to-blue scale whose neutral middle always means no change. The coverage map can be laid over the orthomosaic to see where the flight was thin.

Measuring

The Map Tools button on the map opens four measuring modes. Results are shown in feet, acres, cubic yards and US tons unless you have chosen metric in your account settings; printed reports show both.

Volumes, stockpiles and cut-and-fill

A volume is worked out cell by cell over the surface model at its full resolution: material above the base is cut, space below it is fill, and the difference is the net volume. The base is the surface the pile sits on, and choosing it well matters more than anything else:

• Every volume comes with its uncertainty. Cut, fill and net are each shown with a plus-or-minus figure derived from the map's resolution, alongside the share of the area with missing data.
• Warnings in full. If the boundary seems to cut across the pile, or too much of the area has no data, the result says so.
• Tonnage. Pick a material (sand, gravel, crushed stone, topsoil, clay, riprap, asphalt millings, demolished concrete, coal, wood chips) for a nominal tonnage, or enter a density your site has tested, in lb/yd³ (t/m³ if you have chosen metric). Say whether the material is in place, loose or compacted, and apply a swell or shrink factor.
• Saved and repeated. Save a measurement and attach it to the site, and the same boundary is measured on every survey of that site that covers it, including future flights, giving a volume-over-time chart. Every change to a saved measurement is recorded in its custody history with the old and new values, and every earlier result the change re-measures is recorded with the figures it replaced.
• Reports. A printable volume report (save it as PDF from the print dialog) and a spreadsheet file of the results.

Contours and terrain analysis

Open an elevation model's terrain analysis to make, on demand, any of:

• Contours: lines of equal height, every 1 ft by default with every fifth line drawn heavier; the interval can be set in feet or metres.
• Slope, as a percentage grade or in degrees, and aspect, the direction the ground faces.
• Ponding: low spots where water would collect, from 1 in deep and 100 ft² by default. Click one for its depth, area and volume in cubic yards and gallons.
• Drainage: the paths water would take, for catchments of 1 acre or more by default.

The terrain model (bare ground) is used by default, and the surface model can be chosen instead. Slope and aspect are saved as map images, and contours, ponds and drainage as GeoJSON (a standard format for map lines and shapes that GIS software opens). All are filed beside the stitch with a custody record and become layers in the Site Navigator. Each result states the height reference it is measured from, or says plainly that it is not known.

Analysing

Vegetation indices (Band Analytics)

A multispectral stitch delivers NDVI and NDRE. Band Analytics, on the folder of its bands, computes any other index from the same bands. It has eighteen built in, among them NDVI, GNDVI, NDRE, EVI, EVI2, SAVI, OSAVI, MSAVI2, CI-RedEdge, CI-Green, MTCI, PSRI, VARI, GLI and NDWI (for water), or you can write your own formula from the band names. Preview it live on the map at a resolution you choose, set a colour scheme and value range, and save the result. A saved index is filed beside the stitch, ready to measure from, and appears as a layer in the Site Navigator.

Zonal analytics (agricultural sites)

Zonal analytics turns index maps into numbers per field or block. The zones are the polygons drawn for the site. For each zone and each flight it reports the mean, median, spread, uniformity and range of the chosen index, counting only vegetation by default, and then follows each zone over time: its trend, its season total, and how it compares with the rest of the field. Once an index is tracked on a site, each new flight of that site is added automatically when its stitch finishes. A crop water stress map made from a thermal flight (see "Thermal tools") can be tracked the same way. Flights without radiometric calibration are left out of absolute trends, because their values cannot be compared across flights; they are kept in relative ones. The exception is crop water stress measured against a crop's published limits, whose thermal flights count in its absolute trend; a relative stress map, scaled within its own flight, is kept apart from it and shown in relative views only.

Thermal tools

On a thermal stitch, read the temperature at a point (°F with °C), get statistics for an area, draw isotherms (the areas above, below or between temperatures you set), find hotspots automatically, with a preset for solar sites, and map crop water stress. Hotspots can be turned into findings. The companion paper, Temperatures you can measure, describes these tools and what the temperatures can and cannot claim.

AI detection on the orthomosaic

Run AI on an orthomosaic to find and outline objects in it, from a list you type or a ready-made pack (general, solar PV, roof and property features, construction site, powerline corridor, storm and flood response, livestock and pasture, site monitoring). A large orthomosaic is read tile by tile at full resolution, so nothing is lost by shrinking it first. On a map-aligned orthomosaic the outlines are drawn on the map as a layer, one colour per kind of object. AI detection is priced in credits, shown before you start. Between two dates, AI change lists the objects that appeared, disappeared or stayed (see "Comparing dates").

Findings

A finding is a recorded issue at a place on the site, with a severity and a stage (untriaged, active, resolved or dismissed). Create one at a point or an area in the Site Navigator, or promote AI detections, thermal hotspots and change alerts into findings. Findings can be exported as GeoJSON, CSV or KML. The companion paper, Findings you can act on, covers them in full.

Crop

Crop clips a finished stitch to a shape you draw, such as one field of a farm or one parcel of a development. It makes trimmed copies of the orthomosaic, the elevation models, the index maps, the thermal map and the coverage map in a new folder, leaving the originals untouched. Pixels are kept when their centre falls inside the shape. Cropping is free; 3D models are not cropped.

Ask Orion

Orion, the Omniris assistant, answers questions about your own stitch jobs in plain language: which sites were flown and when, how jobs went and what they cost. On a finished stitch it can also work out the heights of objects (the surface model minus the terrain model), map index zones, summarise temperatures and hotspots, and look at a location on the orthomosaic. It sees only your own data.

Comparing dates

A site flown more than once can be compared date against date. The Compare section of the Site Navigator offers three ways:

The same elevation comparison is offered from the Volume tool (Compare surveys), and a saved volume measurement can compare any two of its flights over its boundary. The two surveys do not need the same resolution or coordinate system: the earlier one is resampled onto the later one's grid. Colour orthomosaics are not compared by value, because differences in lighting between two flights would swamp real change; compare them visually instead. Thermal maps are compared visually too: a computed temperature difference between two flights is not offered. A crop water stress map, where one has been made on both dates, is compared like a vegetation index.

• Reports. A comparison can be printed or downloaded as a PDF report, or as a spreadsheet file, from the web application. The colour map of the change is saved as a layer.
• Change alerts. From a comparison on a site, Alert me sets up a standing watch: when a new flight changes the net volume, the material added or removed, the average index, or the share of the area that changed beyond a threshold you set, in the direction you choose, you and the people you name are told. Alerts are checked every six hours.
• AI change. When both dates have AI detection results from the same model and the same list of objects, AI change lists the objects that appeared, disappeared or stayed between them, and records the appearances and disappearances as findings. It refuses dates whose resolutions differ by more than two times, because the detector finds far more at the finer resolution, which would show objects as "appeared" that were there all along.
• Situation report. One PDF for a site and two dates, in English or Polish: the area, the two dates side by side, the elevation change and AI change, the findings, the sources with verifiable certificates for each file, and the method and its limits. If the elevation could not be compared, the report says why instead of failing.

Sharing, exporting and evidence

The stitch report

Every finished stitch files a stitch report in its results folder, as a self-contained web page and as a PDF, and the completion email links to it. It covers:

• The flight: the site and customer, how many photos were submitted and used, the capture window, the camera, the resolution, the area covered in acres (and hectares), and the flight path over a map.
• The results: the orthomosaic, the elevation models, the multispectral bands and index maps, and the thermal map with any anomalies found, each as a picture.
• How well it was captured: how many photos saw each part of the site, how the multispectral sensor was calibrated, the weather during the flight, and the photos left out before processing and why.
• How well it was processed: the quality grades, the ground-control errors point by point, and the results of the pre-flight check.
• What it delivered, and its limits, in plain sentences, such as the absence of ground control.

Refresh Report brings it up to date after a re-stitch or new flights in the folder. A report under a legal hold is kept, not replaced, and you are told why.

Sharing

Results can also be sent to a Procore project, or to Autodesk Forma, where those integrations are connected; see From design to measured progress. A finished stitch's customer or site name can be changed later, and every result is re-tagged to match.

Downloads and exports

Evidence

Stitched results are evidence in the same vault as your photos, not files in a separate system:

• Custody. Every result is recorded in the tamper-evident custody record as an output of the job that made it, with its fingerprint (a SHA-256 hash). Re-stitching keeps the same record and logs the replacement. Every download, export and share is recorded too, and so is every file a recipient opens through a public link. Results can be covered by a signed custody certificate, which anyone can check on a public verification page.
• Legal holds. Results under a legal hold or retention policy are never replaced. A re-stitch that would overwrite them is refused before anything is charged, with the reason; choosing Back up instead keeps the protected results as they are and writes the new ones to a fresh folder.
• Encryption. Photos in an encrypted folder can be stitched: they are decrypted for the job, and the results are encrypted as they arrive in a protected folder.

The companion paper, Evidence you can prove, describes the custody record, certificates, holds and encryption in full.

Where it runs

Stitching and every tool in this paper are part of Omniris wherever it runs. The same stitching software runs on each.

An on-premises installation that is connected to the Omniris cloud can also send a single job there, by ticking Run in Omniris Cloud in the stitch dialog; it is billed to the organization's Omniris cloud credits.

What the results claim, and what they don't

A map is only as useful as the honesty of its limits.

• A map needs photos taken looking down. Photos more than 25° from straight down are left out of a map, and a flight shot entirely at an angle cannot make one. Such a flight can still make a 3D model.
• Resolution cannot exceed the flight. The finest resolution on offer is set by the height flown and the camera. When the drone recorded only its height above the take-off point, that resolution is an estimate, and the dialog asks you to confirm the real height.
• Without ground control, absolute accuracy is not stated. The map's positions then come from the drone's own GPS. Distances, areas and volumes within the map rely on its relative accuracy; its position on the earth, and comparisons with other flights or designs, are only as good as that GPS.
• Control-point errors are not accuracy. They show how well the map fits the points it was fitted to. Only checkpoints, held out of the fit, measure accuracy elsewhere on the site; a stitch with no checkpoints has no measured accuracy, and the accuracy class can be no better than the least accurate point used.
• Multispectral values depend on calibration. Camera-only calibration is reliable within one flight but not comparable across flights; the sunlight sensor makes flights comparable relative to each other; a reflectance panel puts them on an absolute scale. When a multispectral flight also produces a colour map, the two are built separately and can be offset from each other; the results show the measured offset and say when it is large enough to matter for precise work.
• A volume is only as good as its base. The base is a choice, and a poor one (a lowest point in a rut, a boundary that cuts across the pile) gives a wrong number; the tool warns about the cases it can detect.
• Comparisons depend on alignment. An offset between two flights looks like change. Each comparison is graded for alignment and an unaligned one is labelled an estimate.
• No time is promised. Processing time depends on the job, and Omniris does not quote one in advance.
• Pictures are not maps. A JPG or PNG export is for viewing; it is not map-aligned, and an exported elevation model is a grey picture, not heights.
• Some tools have conditions. Zonal analytics is for agricultural sites. Crop does not work on a stitch in an encrypted folder, and does not crop 3D models. Comparison reports and spreadsheets are produced from the web application; the desktop application shows the comparison itself.
• Mobile. A stitch can be started from the mobile app, with a simpler version of the pre-flight check. The flight-group picker and the offer to build a 3D model instead are on the web and in the desktop application.

Where it matters

Standards

Inputs and results use open, published formats, so your data is never locked into Omniris.

Glossary

Absolute accuracy

How closely positions on the map match their true positions on the earth.

Checkpoint

A surveyed point left out of the fit and used only to measure the map; its error is the map's measured accuracy there.

Cloud Optimized GeoTIFF (COG)

A map-image file laid out so a viewer can read only the part on screen, at the detail needed.

Forward and side overlap

How much each photo shares with the next one along a flight line, and with the photos on the neighbouring line.

Ground control point (GCP)

A marker on the ground whose position has been surveyed, marked in the photos to pin the map to the earth.

Ground reference point

A point taken from a recognisable feature on satellite imagery, used when nothing was surveyed; accurate to about a metre.

Ground sample distance (GSD)

How much ground one pixel covers; the map's resolution.

LAZ

The standard compressed file format for point clouds.

Nadir and oblique

A nadir photo is taken looking straight down; an oblique photo is taken at an angle.

NDVI and NDRE

Standard measures of plant health, from how strongly leaves reflect red, red-edge and near-infrared light.

Orthomosaic

One map-aligned image made from many overlapping photos, corrected so that every pixel is in its true position.

Photogrammetry

Measuring shapes and positions from overlapping photographs.

Point cloud

Millions of measured points on the surfaces the camera saw, each with a position and height.

Reflectance panel

A card of known brightness photographed before a multispectral flight, so the sensor's readings can be put on an absolute scale.

Relative accuracy

How true the map is to itself: distances, areas and volumes measured within it.

RTK

Real-time kinematic positioning: corrections that make the drone's satellite position far more precise.

STAC

SpatioTemporal Asset Catalog, an open standard for describing collections of map data.

Surface model (DSM)

A height map of the highest thing at each spot: ground, vegetation or structure.

Terrain model (DTM)

A height map of the bare ground, with vegetation and structures filtered out.