DJI Terra License Guide
Confirm the software edition, license term and authorization mode before redeeming an activation code. Match the entitlement to the intended devices and processing workflow.
Survey quality guide | Drone Emotions
DJI Terra Accuracy, GCP & RTK Guide
Understand how to plan, control and verify the accuracy of your DJI Terra maps, models and point clouds. Learn how GCPs, independent check points, RTK, PPK and coordinate systems work together, then use a clear workflow to assess the finished deliverables.
How to achieve reliable DJI Terra survey results
Accuracy starts with the survey specification and continues through acquisition, reconstruction and delivery. A successful processing run is one stage of that sequence.
Define what must be accurate
List the required deliverables, horizontal and vertical tolerances, coordinate reference and measurement purpose. A visual inspection model and an engineering surface may need different acceptance checks.
Agree on the evidence to deliver before planning capture or processing.
Control the reconstruction
Use suitable image or LiDAR data, reliable positioning and correctly surveyed control. Keep the coordinate definitions and source records with the project.
Review input quality before trying to improve a weak dataset through settings alone.
Verify with independent evidence
Reserve surveyed points for checking rather than adjustment. Assess the exported product as well as the processing report.
Report the observed errors, point coverage and test limitations alongside the result.
GSD, relative accuracy and absolute accuracy
Keep resolution, image residuals and surveyed position errors separate when reviewing a map or 3D model. Each answers a different question.
| Term | What it describes | How to use it |
|---|---|---|
| GSD — Ground Sample Distance | Ground distance represented by one image pixel | Plan the detail you need to see; then verify positional accuracy separately |
| Relative accuracy | Agreement of distances and geometry within the model | Check dimensions and local consistency for the intended measurement |
| Absolute accuracy | Agreement with an external survey reference | Compare independent surveyed positions in the same coordinate system |
| Horizontal accuracy | Position in the horizontal plane | Review easting and northing errors together |
| Vertical accuracy | Agreement of heights or elevations | Check the height reference and the surface being measured |
| Reprojection error | Difference between image marks and projected positions, in pixels | Inspect marking and image consistency |
| RMSE | Root mean square of a defined set of errors | Identify the reference, units, point roles and error components |
On smaller screens, swipe the table horizontally to see every column.
Example: a GSD of 2 cm/pixel describes image sampling. It does not establish a 2 cm horizontal or vertical error. Similarly, a low reprojection error describes image agreement and needs separate ground verification.
Reference: DJI definitions of GSD and GCP error.
RTK vs PPK: improve positioning, then verify the output
Accurate camera or trajectory positioning can reduce the amount of control needed by a suitable workflow. Keep its verification requirements separate from its positioning method.
| Method | When it acts | What to prepare |
|---|---|---|
| RTK — Real-Time Kinematic | Corrects GNSS positioning during acquisition | A suitable correction source, reliable observations and documented base/reference coordinates |
| PPK — Post-Processed Kinematic | Calculates corrected positioning after acquisition | Compatible aircraft observations, corresponding base data and timing records |
| Ground control | Constrains reconstruction using surveyed locations | Visible targets or appropriate LiDAR control, accurate survey coordinates and defined point roles |
On smaller screens, swipe the table horizontally to see every column.
Reference: DJI explanation of RTK and PPK principles.
Review the actual RTK record
A FIX solution is useful input information. Review its availability across the acquisition, the correction source and the position used for a local base.
A base coordinate error can affect the reference of the entire dataset. Keep independent ground checks in the verification plan even when RTK performs well.
Prepare PPK before the flight
Confirm that your equipment records the observations and event information required by the selected PPK workflow. Preserve original folders and matching base-station records.
Check supported inputs for the specific Terra project. PPK does not recover missing imagery or remove blur and occlusion.
For native Zenmuse datasets, consult the separate LiDAR POS and PPK import instructions. Describe the sensor and processing route when requesting advice; the available positioning options depend on that setup.
GCPs and check points: control and verification
A surveyed target can become either a control point or a check point. Its role in processing determines how you should interpret its reported error.
| Point type | Used in adjustment? | Purpose |
|---|---|---|
| Control point — GCP | Yes | Constrains the aerotriangulation result using surveyed coordinates |
| Check point | No | Evaluates agreement with surveyed coordinates withheld from adjustment |
On smaller screens, swipe the table horizontally to see every column.
Reference: Current DJI GCP and constraint manual.
Decide which points will control the reconstruction and which will test it. Use stable point IDs and record the split. Review the final point types after import and before optimization.
Keep the check set independent. A point used to fit or repeatedly tune the solution has a different evidential role from a point reserved for assessment. If you change the control plan after reviewing errors, document the change and preserve a suitable independent check set.
Low control-point residuals show how the solution fits its constraints. They can be helpful for finding mistakes, but they do not describe accuracy everywhere outside those points. Examine the distribution and errors of the check points as well.
Being withheld from adjustment does not remove a shared survey-reference error. If photos and check points use the same incorrectly positioned base, their agreement can conceal an absolute offset. Verify the base and reference against suitable external survey evidence.
How to plan GCPs and independent check points
Point visibility, survey quality and distribution all affect the result. A point count alone does not define a complete control or verification plan.
Plan coverage and geometry
Spread points across the relevant survey area. Include interior coverage and locations that reveal changes in terrain and geometry. Avoid relying on a cluster at the take-off location or points along one line.
For corridors or irregular sites, plan where independent checks will test the portions of the deliverable that matter to the client.
Choose clear, stable targets
Use points that can be identified consistently in photographs and measured reliably on the ground. Provide a point ID, site photograph and exact marking description.
Keep targets unobstructed and choose locations that remain stable throughout acquisition. Size them for the intended image resolution and viewing geometry.
Record survey quality
Keep the measurement method, horizontal and vertical reference, units and estimated uncertainty with every point set.
Inspect repeat observations and unexpected height differences before importing. A precisely marked target still depends on the quality of its surveyed coordinate.
Current DJI guidance: the GCP manual recommends at least five evenly distributed GCPs and marking eight to nine images from different angles per point. Its photogrammetry accuracy guide also recommends at least five GCPs and five check points for an actual project. Treat these as DJI planning guidance; adapt the survey design and verification coverage to your site and acceptance requirements.
Reference: DJI current quantity and marking recommendations.
Reference: DJI photogrammetry accuracy-check planning.
Coordinate systems, EPSG codes, geoids and heights
Write down the references used by the photos or trajectory, the surveyed points and the deliverables. Confirm how they relate before applying constraints or evaluating errors.
Photo or trajectory POS
Identify the reference of the recorded positioning data. Keep the correction-service or base information and describe any coordinate conversion performed before import.
Do not change the declared reference simply to make a dataset appear aligned.
Surveyed ground coordinates
Select the coordinate system in which the control and check points were actually measured or transformed. Confirm axis order, units, projection zone and height reference.
Check a known point in the import preview before trusting the whole file.
Output coordinates
Select the required delivery reference and review the exported result. The current manual says that, after marking GCPs, the output coordinate system must be consistent with their coordinate system.
Record the definition used for delivery, including any PRJ or transformation file.
Ellipsoidal height and local elevation
GNSS ellipsoidal heights and orthometric elevations use different references. Confirm the intended height system with the survey data and the client’s project specification. A horizontal EPSG selection alone may not define the required vertical reference.
DJI’s current documentation includes selection of vertical coordinate systems and support for importing local geoid files. Verify the applicable model, area of coverage and input/output reference before using that conversion.
Reference: DJI local geoid import announcement.
Investigate a consistent height shift before adjusting it. Check the base coordinate, antenna-height handling, source heights, geoid and output units. Keep any correction traceable to an identified reference or survey observation.
Import GCP coordinates without column or unit mistakes
A readable file is only the start. Check that every imported value still represents the intended point, coordinate component and measurement reference.
| Field or setting | What to confirm | Typical issue to catch |
|---|---|---|
| Point ID | Unique name matching the field record | Duplicate or exchanged target names |
| Coordinate columns | Correct mapping of geographic or projected coordinates | Swapped latitude/longitude or easting/northing |
| Height column | Correct surveyed height and reference | Ellipsoidal and local elevations mixed |
| File parsing | Delimiter, decimal separator and header rows | Columns split incorrectly or a header read as a point |
| Accuracy fields | Measured horizontal and vertical uncertainty when using custom accuracy | An unrealistically small value used as a target tolerance |
On smaller screens, swipe the table horizontally to see every column.
Terra supports TXT and CSV ground-point files. Configure the parser and map each column using the preview. For projected data, check that X/E means easting and Y/N means northing in the Terra mapping.
If you use Auto GCP in an arbitrary coordinate system, the current manual requires corresponding WGS84 coordinates to assist positioning. Review that scenario before preparing the import file.
Reference: DJI current ground-point file and import requirements.
Keep the original survey export and a separate processing copy. Record the parsing and coordinate settings with the project so the import can be reproduced.
DJI Terra GCP workflow: from photos to verified results
This is an original working sequence combining input checks, current manual terminology and delivery review. Consult the linked DJI instructions for the full interface procedure.
-
Inspect original data and create the project
Import the selected photographs. Review their sharpness, coverage, positioning and camera information. Keep the equipment, release and original-data inventory with the project record.
-
Run initial aerotriangulation
Calculate the initial camera solution and inspect its coverage. The current manual recommends performing aerotriangulation before marking to improve predicted point locations.
-
Open GCPs and Constraints and import the points
Under Preprocessing (AT), open GCPs and Constraints. Import the ground-point file, confirm its coordinate properties and preview the column mapping.
-
Mark and review target positions
Identify the same surveyed target centre in each selected photograph. Review automatic or assisted marks individually, especially near image edges or on ambiguous features. Follow the current marking guidance above.
-
Confirm point roles and adjustment settings
Set the reserved check points to the check role. Review coordinate consistency, measurement accuracy and the Image POS constraint. Keep a record of the selected settings.
-
Optimize, reconstruct and verify
Optimize the marked solution using the appropriate interface for the installed release. Review the report, generate the required products and test their exported coordinates and surfaces against independent evidence.
Reference: DJI GCP processing sequence.
Use the current interface for your installed release. Older instructions may show “GCP Management” where newer documentation uses “GCPs and Constraints.” Preserve the point roles and adjustment logic when following the equivalent controls.
GCP accuracy settings and the Image POS constraint
Weights describe confidence in input observations. Coordinate consistency determines whether different observations can constrain the same solution meaningfully.
Ground-point accuracy is a weight
Use an accuracy estimate supported by the ground survey. Smaller uncertainty values give the point stronger influence in adjustment; they do not improve the observation itself.
Separate a client’s required tolerance from the uncertainty of the coordinates you measured. Keep any custom weighting decision with the processing notes.
Image POS must fit the chosen reference
The current manual recommends enabling Constrain with Image POS Data when the POS and ground-control coordinate and elevation references are consistent. It specifies disabling the option for its local/arbitrary-coordinate scenarios.
For other height or transformation cases, follow the current scenario-specific manual rather than applying one setting to every project.
If a point has a large error, first check its name, measurement, mark and reference. Changing weights or deleting the point should follow an identified reason. Record the issue and its resolution so later comparisons remain meaningful.
How to interpret the DJI Terra quality report
Identify the units, reference observations and point roles behind every number. Low residuals in one category do not automatically establish the accuracy of a different category.
| Report item | What it helps assess | What to verify next |
|---|---|---|
| Calibrated images and coverage | Whether the intended photographs contributed to the solution | Look for spatial gaps, rejected groups and weak coverage |
| Reprojection error / RMS | Image-coordinate agreement, reported in pixels | Review marking, imaging and geometry |
| Georeferencing RMSE | Agreement of calculated camera positions with recorded POS | Check POS quality and reference; then use ground check points |
| Control-point residuals | Fit to surveyed points used as constraints | Inspect individual errors and weighting |
| Check-point errors | Agreement at surveyed locations withheld from adjustment | Review components, distribution and output-specific checks |
On smaller screens, swipe the table horizontally to see every column.
Reference: DJI aerotriangulation accuracy-check method.
Read the per-point values as well as the summary. Check whether the largest errors occur at edges, in one flight block, at a particular elevation or around weakly reconstructed features.
A missing statistic or an unused check point is a missing piece of evidence, not a zero error. Keep the report with the software release, inputs, control split and final settings.
Reference: DJI quality-report documentation.
Read the complete DJI Terra Quality Report Guide
Continue with our DJI Terra Quality Report Guide for a review of the main report indicators, a worked RMSE example, separate 2D, 3D and LiDAR checks, and a practical delivery checklist.
Use it alongside your accuracy plan to connect the reported numbers with the exported results and the verification evidence required by your project.
RMSE explained with a simple worked example
RMSE summarizes a defined set of errors. Its usefulness depends on the reference quality, independent observations and coverage of the tested deliverable.
For independent check points, define each error as the output coordinate minus the surveyed reference coordinate. Use the same coordinate system and units. In a projected metric system, write the differences as dE, dN and dZ.
Horizontal RMSE: square root of the mean of dE² + dN².
Vertical RMSE: square root of the mean of dZ².
Vertical bias: mean of the signed dZ values.
| Check point | Output − reference | Squared error |
|---|---|---|
| CP-A | +0.020 m | 0.0004 m² |
| CP-B | −0.020 m | 0.0004 m² |
| CP-C | +0.040 m | 0.0016 m² |
| CP-D | −0.040 m | 0.0016 m² |
On smaller screens, swipe the table horizontally to see every column.
Calculation: the mean squared error is 0.0010 m². Its square root is approximately 0.0316 m, or 3.16 cm. The signed mean is zero, so positive and negative errors cancel in the bias while the RMSE remains non-zero.
These four invented values explain the calculation. They are not a DJI benchmark, a recommended check-point count or proof of a survey accuracy class. For a real report, include all applicable observations, the point distribution, uncertainty and any documented exclusions.
Report a set of results. Include horizontal and vertical errors, signed bias, largest observed discrepancies and the number of tested points. Describe the tested surface and locations. Use the acceptance method agreed for the project when deciding whether the deliverable meets its requirements.
LiDAR accuracy control for Zenmuse workflows
Photogrammetric GCP marking and LiDAR elevation control use different mechanisms. Assess the reported result within the scope of the actual Terra function.
Control elevation in the LiDAR workflow
The current manual describes imported control points that adjust local point-cloud elevation using their planar positions. These points do not require the photo marking used in RGB aerotriangulation.
Select the correct input reference and inspect the ground around each point. Keep an independent set for checking the resulting elevations.
Select a suitable check surface
DJI recommends flat, open areas, clear surroundings and no overhead obstruction for this function. The recorded reference should represent the surface sampled by the point-cloud check.
Review unexpected errors near vegetation, steps, edges or steep changes before interpreting them as trajectory or coordinate problems.
Inspect strips and fine detail
Review cross-sections and overlapping strips for offsets or excessive thickness. Evaluate optimization settings against the final measurement purpose.
DJI’s processing guidance cautions against smoothing when small elevation changes must be preserved. Keep the unsmoothed result for comparison when assessing fine features.
Scope of the check: this local elevation-control function does not establish complete horizontal or XYZ accuracy. Plan additional horizontal and output-specific checks when those components form part of the specification.
Reference: DJI LiDAR accuracy-check methods.
For sensor and native-data requirements, consult the supported drones and sensors guide. Preserve complete raw flight folders and identify the payload used before comparing LiDAR projects.
Verify the exported map, surface and 3D deliverables
The aerotriangulation report describes the underlying solution. Delivery review also needs to examine the product and processing choices the client will actually use.
Orthomosaic / DOM
Inspect clearly identifiable surveyed locations in the exported map. Confirm its coordinate reference, units, coverage and behaviour in the receiving GIS.
Look for seam distortions and displaced features in the portions of the map that will be used for measurement.
DSM, DEM and point cloud
Compare elevations on the appropriate surface. A surface model can include roofs and vegetation; a ground model depends on its ground classification.
For a slope, ensure that the surveyed reference and the sampled output location correspond. Record interpolation or sampling choices.
Mesh and downstream measurements
Review representative dimensions, boundaries and geometry in the exported model. Document simplification, editing or other delivery changes.
For volume work, record the reference/base surface and boundary together with the point-cloud or mesh quality checks.
DJI’s accuracy guide recommends reviewing the finished raster products in GIS in addition to the aerotriangulation report. Apply the same output-first principle to the client’s point-cloud, mesh and measurement deliverables.
Reference: DJI raster output accuracy-check guidance.
Troubleshoot shifts, residuals and reconstruction errors
Use the pattern of the error to choose the next investigation. Each row gives possible causes; confirm them from the data before applying a correction.
| Observed issue | Possible cause | Check before changing settings |
|---|---|---|
| A broadly consistent vertical shift | Different height references or a base/setup error | Source and output heights, geoid, base coordinate and antenna-height handling |
| Large horizontal displacement | Wrong projection, reference, zone, units or column mapping | A known point in the import and export coordinate definitions |
| Good control fit but weak check points | Uneven control, an incorrect mark or an input/reference problem | Independent point roles, coverage and individual observations |
| Large reprojection error at one target | Wrong centre, ambiguous feature, blur or edge marking | The original photographs and matching target ID |
| Gaps, bending or local distortion | Weak image geometry, overlap, visibility or calibration | Spatial coverage, aligned image groups and camera information |
| Layered LiDAR strips or noisy surfaces | Trajectory/calibration issues, weak overlap or unsuitable settings | Complete source data, cross-sections and relevant DJI processing instructions |
On smaller screens, swipe the table horizontally to see every column.
Reference: DJI guidance for excessive reconstruction error.
Preserve the baseline before testing a change. Write down the suspected cause, change one relevant factor and compare the new result against the same independent checks. If the cause remains unclear, send the original dataset description, settings and report through our contact page.
A practical survey-accuracy checklist
Keep the project specification, field records and processing evidence together so that the result can be reviewed and reproduced.
Before acquisition
- Required products and horizontal/vertical acceptance criteria.
- Coordinate and elevation references, units and downstream software.
- Camera or payload, positioning method and preserved observation files.
- Control and check-point distribution, IDs and measurement quality.
Before accepting the result
- Original-data inventory and software release recorded.
- Correct coordinate import, point roles and adjustment settings.
- Individual check errors and report statistics reviewed.
- Final exported products tested in their delivery workflow.
- Exclusions, unresolved issues and verification coverage documented.
For repeat surveys, preserve the reference definitions and verification method across dates. Evaluate whether observed changes exceed the combined uncertainties of the compared results before interpreting small differences.
DJI Terra accuracy, GCP and RTK FAQ
Open a question for a direct answer. The earlier sections provide the workflow, definitions and official references.
Does RTK remove the need for ground check points?
Keep an independent verification plan. RTK improves the positioning input, while check points assess the reconstructed result against surveyed locations. The control requirement and the verification requirement are separate decisions.
Reference: Supporting DJI documentation.
Is GSD the same as map accuracy?
GSD describes image sampling on the ground. Positional accuracy needs its own assessment against a defined reference. A 2 cm/pixel dataset does not by itself demonstrate a 2 cm mapping error.
What is the difference between a GCP and a check point?
A control point constrains the reconstruction. A check point is withheld from adjustment and used to evaluate agreement with its surveyed coordinate. Confirm the assigned roles before optimization.
Reference: Supporting DJI documentation.
How many GCPs should I use?
The current GCP manual recommends at least five evenly distributed control points. Its accuracy guide also recommends a separate check-point set. Adapt coverage and verification to the site and agreed acceptance method rather than treating a minimum as a complete survey design.
Reference: Supporting DJI documentation.
How many photographs should I mark for each point?
The current manual recommends eight to nine images from different angles for each GCP. Review image quality and individual marks, including any generated automatically.
Reference: Supporting DJI documentation.
Should I always enable Constrain with Image POS Data?
Choose it for the coordinate scenario documented in the current manual. Consistent POS and ground references can be used together; local or arbitrary scenarios have specific instructions. Record the selected setting.
Reference: Supporting DJI documentation.
Does a low reprojection error prove centimetre accuracy?
It describes image-coordinate agreement in pixels. Use separate ground checks and inspect the exported product to assess horizontal and vertical accuracy.
Can a zero mean error still have a non-zero RMSE?
Yes. Positive and negative errors can cancel in the signed mean. RMSE squares the errors before averaging, so their magnitudes still contribute. The worked example above shows this.
Why is my model consistently too high or too low?
Investigate source and output height references, geoid settings, the base coordinate and observation handling. A consistent shift suggests a reference or setup issue worth checking before applying an offset.
Do LiDAR control points use the same marking procedure as photographs?
The current LiDAR function uses planar point locations for local elevation control and checking without manual photo marking. Review its documented scope and plan other accuracy components separately.
Reference: Supporting DJI documentation.
Can PPK recover a flight with poor images?
PPK addresses positioning when the required observations are available. Blur, missing views, weak geometry and occluded subjects need their own assessment and may require new acquisition.
What should I send when asking for accuracy support?
Send the equipment and Terra release, source-data description, positioning method, coordinate and height references, control/check split, intended deliverables and report. Include the error pattern and any processing changes already tested.
For other software questions, explore the complete DJI Terra FAQ.
Continue planning the complete DJI Terra workflow
Match the equipment and processing setup
Check the supported data route and prepare a workstation for the intended project size.
Prepare Terra and its tools
Review the edition, installation and reconstruction functions required by the deliverables.
Keep the official references available
Use DJI’s documentation for your release and keep its report guidance with the project.
Plan your accuracy and verification requirements
Tell us what you need to deliver, the equipment and positioning method you use, and the error you are investigating. Include the project reference and report so that we can discuss the next useful checks.
Original educational guide prepared by Drone Emotions. Documentation reviewed on 2 October 2026. Current manual procedures take precedence over older interface descriptions. Accuracy acceptance should follow the project specification and the evidence collected for the actual deliverables. About Drone Emotions.