WORKFLOWS / KNOWLEDGE & DESIGN
Surveying, Point Clouds & Construction Management | Separating looks from accuracy
A beautiful 3D model and data you can measure with are not the same thing. Build the job from the required accuracy and the verification method.
Read this page in order
Work backward from the required deliverables
Orthomosaics, point clouds, 3D models, volume comparisons. For each use, line up the coordinate system, the required accuracy, the area covered, the delivery format and the acceptance method. If the results will be used for public surveys or construction records, check the client's specification and the applicable standards separately. This page does not certify surveys or judge compliance.
| Deliverable | What to ask first |
|---|---|
| Orthomosaic | Coordinate system, resolution, tolerance for gaps |
| Point cloud | Density, gaps, horizontal and vertical accuracy |
| Differences such as earthwork volumes | Comparison dates, boundaries, reference surface, calculation method |
| Construction records | How the same spot is tracked over time, capture interval, who verifies |
Separate acquisition, processing and verification
Set up the capture conditions and the positional reference, process the data from the imagery, then evaluate the results against independent check points. Keep these three as separate stages and it becomes far easier to find where an error crept in. High-precision positioning on the equipment does not by itself guarantee the accuracy of the whole deliverable.
- Record the software and settings used for processing, and the version of each input file.
- Keep the points used for verification distinct from the points used for adjustment.
- Make the coordinate system, vertical datum and units mandatory.
- Confirm you are comparing the same area at the same point in time.
Keep a worked example of error calculation
The example below is fictional and is there to make the math clear. The numbers are not tolerances or pass/fail thresholds. Actual acceptance follows the specification that applies to your job.
If the check-point errors are 0.02, -0.03 and 0.01 m:
RMSE = sqrt((0.02² + (-0.03)² + 0.01²) / 3)
≈ 0.0216 m
Always state alongside: number of points / point layout / reference / measurement method
※ Three points alone may not represent the quality of the whole siteA single average can hide a large local error. Keep the per-point errors and their distribution, and note the areas that were not observed.
Turning it into screens and data
Attach the coordinate system, version and verification results to every deliverable shown on the map. Warn before overlaying files in different coordinate systems, and record the transformation parameters. Store the inputs and processing conditions together as one bundle so that any delivered output can be regenerated.
DEFINITION OF DONE
When you build this, check these points.
- A deliverable missing units or a coordinate system can be held for review.
- Adjustment points and check points can be told apart.
- Every deliverable can be traced back to its processing conditions and input files.
Design checklist items for spec reviews and functional testing.
How to read this page
This is our own design proposal. The implementation and numerical examples are not evidence of real-world results or certification. Verify them against the conditions of the work you apply them to.
Read how we make these pages and our editorial policy