Glossary — GPR processing
GPR Migration Processing: Choosing the Right Velocity
The processing step that turns arches back into points, so pipes, slab edges and sloping layers sit where they really are.
Quick AnswerGPR migration processing moves reflected energy back to where the reflector truly is. It collapses each hyperbola to a point at its apex and puts sloping layers at their real angle. The result is a high resolution image that looks like the ground. It depends on a correct velocity: too low leaves arch limbs, too high bends them into smiles. More in the GPR, SUE & Land Survey Glossary.
What is migration in GPR processing?
Migration is a processing step that undoes the spreading caused by the radar's wide beam. Because the receiving antenna sees targets to its side, a point produces a hyperbola and a sloping layer appears shifted and flatter than it is. Migration sums or back-propagates the recorded energy along those geometric paths, collapsing hyperbolas to points and restoring true dips.
Ground penetrating radar (GPR) builds its picture of the subsurface from echoes of electromagnetic waves. Each echo is logged under the antenna, not under the target, and migration corrects that offset.
Which methods are used?
Three families cover almost all GPR migration processing software.
| Method | Idea | Notes |
|---|---|---|
| Kirchhoff (diffraction summation) | Sums amplitude along hyperbolas for every output point | Robust; handles varying velocity reasonably |
| F-K (Stolt) | Works in the frequency-wavenumber domain | Fast; assumes constant velocity |
| Finite-difference | Steps the wavefield down through the ground | Handles layered velocity; slower |
Processing time is part of our per-metre rate, not an extra line; try the rate calculator for your scan length.
Two-dimensional migration assumes the service crosses the line at right angles. For a pipe crossing at an angle, 3-D migration on a grid works better.
Why does velocity matter so much?
Migration uses velocity to predict the hyperbola each buried object should draw. With a velocity that is too low the arch does not fully collapse and its limbs remain; with one that is too high it overshoots and leaves an upward-curving smile. Processors test several velocities and choose the one that focuses arches best, a useful second check on hyperbola fitting.
Where does migration fit in GPR data processing?
Near the end, after time-zero correction, dewow, background removal, gain and filtering, and before depth conversion. Migrating noisy data spreads the noise into curved artefacts, so cleaning comes first. For 3-D migration, the positions of the data collected must be accurate too. Lines are tied to control fixed with GNSS, as our RTK (Real Time Kinematic) Survey guide describes. Corrections may come from a national reference network, so read CORS Network in India for that part.
A worked velocity test by migration
Migrating the same arch at three velocities shows which one is right. Take an apex at 16 ns. Migrated at 0.08 m/ns the limbs remain; at 0.12 m/ns the arch turns into an upward smile; at 0.10 m/ns it collapses to a tight dot. That focused velocity places the target at 0.10 × 16 / 2 = 0.80 m. Choosing 0.12 m/ns by mistake would have reported 0.96 m, so the test is worth doing before any depth leaves the office.
Migration mistakes we see in other people's data
Nearly all of them come down to the wrong velocity or the wrong geometry.
- Borrowed velocity. A value from another line, day or site, with different ground conditions, is applied without checking the arches on this one.
- 2-D migration of an oblique pipe. The method assumes a right-angle crossing; at a skew the result stays blurred.
- Migrating before cleaning. Ringing that was not removed first spreads into curved smears.
- Judging by looks. An image that seems sharper is not necessarily correct; the test is whether known targets sit at their taped depths.
How migration is stated in our deliverables
Whenever a Mezux image has been migrated, its caption says so and gives the velocity used. We also keep the unmigrated section, because that is what the field crew saw on screen while marking. Comparing the two avoids confusion when an arch on site becomes a dot in the report. No geophysical survey standard prescribes a migration algorithm. ASTM D6432-19 treats processing as part of the GPR method in broad terms, leaving the choice to the interpreter, which is why the record of what was done matters.
When is migration worth the time?
When geometry has to be measured, not only marked. On a route where the job is to spray paint over each of the utility lines before excavation, the apex of an unmigrated arch already gives position and depth. Migration adds little there. It pays off in three cases. The first is when closely spaced targets such as rebar or duct banks must be counted and spaced. The second is when sloping layers, trench walls or tunnel crowns must be shown at their true angle. The third is a grid that will be sliced into plan maps, where uncollapsed arches would smear every service into a broad band. In those cases we migrate, test the velocity on known targets and record the settings with the result.
An example on a slab
Before coring a 250 mm slab, the raw high frequency scan shows a row of tight arches from rebar. After migration at the right velocity each bar becomes a small dot, and the gap between two bars can be measured directly. Under-migrated data would make the bars look wider and the gap narrower. Before and after images appear in the Mezux guide: GPR Radargram Library.
For processed data you can design from, ask a survey engineer for your project.
Frequently asked questions
Is migration always applied?
No: for simple marking on site, unmigrated data is enough. Migration matters for exact geometry, slices and 3-D work.
Can migration give the pipe diameter?
It sharpens the image of the subsurface object, which helps, but size from GPR remains an estimate until exposure confirms it.
Will migrated images look different from what I saw on site?
Yes: arches become points. To read both versions, see How to Read a GPR Radargram.
Does migration change the depth of a target?
Not much for a point target; the apex time stays. It mainly moves energy sideways and corrects dipping layers, which can shift their depth noticeably.
Can migration be done on site?
Some field software can, roughly, but careful migration with velocity testing is normally done in the office after the survey.
Why do the edges of a migrated section look curved?
At the start and end of a line there is not enough data on one side, so the process leaves arc-shaped artefacts there, and these are ignored in interpretation.
Is migration the same as filtering?
No: filters remove unwanted frequencies or flat ringing; migration repositions energy according to geometry. They are separate steps in the chain.
Get processed, design-ready GPR data
Written by Saniya Parveen · Technically reviewed by Rahul Rai · Updated 8 November 2026