Comparison guide
Manual, vehicle, or drone: five methods compared.
Five inspection methods compared on cost, closure time, crew size, and the smallest defect each can capture, with cited figures throughout, including the cases where a survey vehicle or a walk-down is still the right call.
Updated: July 2026
At a glance
The master comparison.
Sourced figures are attributed in the cell. Scroll the table sideways on mobile.
| Criterion | Manual walking survey | Windshield / dashcam survey | Specialized survey vehicle (laser/LCMS class) | UAV imagery + manual engineer review | Imagery + HALO AI |
|---|---|---|---|---|---|
| Coverage completeness | Full surface in principle, but sampled and estimated in practice: typically 10 to 20% of a large area is actually inspected. | Drive-lane view only; misses off-lane areas and fine surface detail. | Full lane coverage at speed; strongest on rutting and roughness. | 100% of the surface imaged; review depth limited by engineer time. | 100% of the imaged surface analyzed, every detected defect measured. |
| Survey speed | 2 to 4 hours for a 3 km pass (Fadron published comparison); weeks for large networks. | Fastest coarse pass there is: drives the network with traffic. No sourced throughput figure to publish. | Surveys at or near highway speed; the fastest full-lane instrument method. | Capture takes 20 to 40 minutes per 3 km (Fadron published comparison); engineer review then adds days per site. | 20 to 40 minutes of flying per 3 km (Fadron published comparison), roughly 5 to 30 ha per flight hour at pavement resolution, analysis delivered within 24 hours. |
| Cost | $200 to $300 per km for a traditional survey; a full airport PCI survey runs 200+ inspector-hours and $50,000 to $150,000 (Fadron published comparison; Pavement Inspection Guide). | Lowest cost per pass of any method; no sourced per-km figure to publish. | High capital and mobilization cost, justified on large highway networks; no sourced per-km figure to publish. | Capture cost matches the drone column; expert review hours dominate the total and scale with area. | $90 to $120 per km, versus $200 to $300 traditional; $2,100 versus $6,600 for a 3,000 m runway (Fadron published comparison). |
| Crew required | 4 to 6 inspectors on the surface (Fadron published comparison). | A driver, with or without a rater riding along. | A specialist operator and the survey vehicle, usually scheduled in advance. | A 2-person flight crew, plus the reviewing engineer's time. | 2 personnel for capture (Fadron published comparison); analysis is automated. |
| Consistency / objectivity | Varies inspector to inspector; depends on training and fatigue. | Low; subjective and coarse by design. | High for its measured metrics; instrument-based rut and roughness. | Imagery is consistent, but defect classification still depends on the reviewer. | High; the same algorithm scores every dataset the same way. |
| Smallest defect reliably captured | Whatever the eye resolves up close. | Only large, obvious distress. | Fine cracking and rutting within the sensor specification. | Down to the imagery resolution (GSD) the engineer inspects. | Distresses down to 1 mm at the required imagery resolution. |
| Traffic disruption / closures | High; crews on the surface. A 3,000 m runway survey takes about a 4-hour closure with 5 staff (Fadron published comparison). | None; drives with traffic. | Low; drives at or near traffic speed. | Minimal; a short flight, no crews on the surface. | Minimal; the same 3,000 m runway is flown in about a 40-minute closure with 2 staff (Fadron published comparison). |
| Output standard (D6433 / D5340) | Reference method for ASTM D6433 and D5340. | No formal standard; screening only. | D6433 plus rut and IRI indices. | D6433 / D5340 when the engineer applies it. | ASTM D6433 and D5340, with deduct values and max CDV printed in the report so the math is checkable. |
| Repeatability for change tracking | Limited; hard to reproduce a survey exactly. | Poor. | Good; georeferenced, repeatable runs. | Imagery repeats well; the manual review step varies. | Excellent; the same imagery re-scores identically, change tracking built in. |
| Best suited for | Small sites, disputes, and the reference audit. | Cheap first-pass screening of large networks. | Highway networks needing rutting and IRI at speed. | One-off detailed jobs where scale is not the constraint. | Objective, repeatable ASTM condition scoring at scale, across sites and years. |
Sources: "Fadron published comparison" refers to Fadron's published drone-versus-traditional survey comparison. "Pavement Inspection Guide" refers to the published guide's manual airport PCI survey figures. Cells without an attribution are our qualitative judgment.
Where we lose
No method wins on every row. Including ours.
Specialized vehicles still beat us on rutting and IRI at highway speed. The walking inspector still beats us on the judgment call in a dispute. Where we win is objective, repeatable surface distress scoring across 100% of the surface, to ASTM, at scale, and we would rather you know both before you buy anything from anyone.
Method by method
Where each method genuinely wins.
Manual walking survey
The walking survey needs zero equipment and remains the reference method that every other approach is measured against. An experienced inspector on the surface can judge context that no camera sees, and for a small site or a contested claim it is often the right call. Its honest limit is scale and repeatability: a full manual airport PCI survey runs to 200+ inspector-hours and $50,000 to $150,000 (Pavement Inspection Guide), and two inspectors, or the same inspector on two days, will not produce identical results across a large area.
Windshield / dashcam survey
A windshield survey is the cheapest coarse screen there is. Mount a camera, drive the network with traffic, and you get a fast, low-cost first pass that flags where the worst pavement is. It is genuinely useful for triage. What it cannot do is measure: it sees only the drive lane, misses fine distress, and carries no formal standard, so it screens rather than scores.
Specialized survey vehicle (laser / LCMS class)
This is the strongest tool for what it is built to do. Laser and LCMS-class vehicles excel at rutting and IRI roughness at highway speed, with instrument-grade, repeatable, georeferenced output, and for a highway agency running a large paved network that is exactly the data they need. The trade-offs are capital cost, availability, and specialization: the equipment is expensive and scheduled, and it is built for road geometry rather than parking lots, aprons, or structures.
UAV imagery + manual engineer review
Flying a drone and having a civil engineer review the imagery produces genuinely high-quality output: complete surface coverage, defensible classification, and standards-based scoring when the engineer applies it. The concession is that it does not scale. Every project is bottlenecked on expert review time, so cost and turnaround climb with area, which makes it best for one-off detailed jobs rather than recurring, portfolio-wide surveys.
Imagery + HALO AI
HALO AI takes the same imagery a UAV workflow produces, from any survey-grade drone, and analyzes 100% of the surface automatically, with the same algorithm applied every time. What comes back is an ASTM D6433 or D5340 PCI with the deduct values and max CDV printed, so the score is checkable. That is where it wins, with change tracking between surveys built in. On detection performance the evidence we cite is not ours: research on this class of crack-detection model reports recall around 95% (FAA Tech Center, 2023). It does not replace engineering judgment, and where the score has to carry a signature, the Engineer-Verified QA add-on puts a civil engineer's review and sign-off behind the classifications and the PCI math, through the Early Access program.
Independent evidence
Drone survey at network scale, from third parties.
Two published examples of drone-based road survey at scale, both independent third-party programs.
Kuwait Public Works Ministry, 2024
Mapped 1,000 km of major roads by drone in six weeks, at a reported 40% lower cost than conventional survey.
City of Harrisonburg, Virginia, 2021
Covered 350 km of city streets for potholes in three days using drone survey.
The economics
The survey is cheap. The waiting is not.
The survey cost matters less than it looks, because the money is in what the survey lets you do next. Early repairs run 4 to 5 times cheaper than delayed rehabilitation (EIC deck), and a deferred bill compounds. Whichever method you choose, the expensive mistake is surveying too coarsely to catch distress while it is still cheap to fix.
Not either / or
Combining methods.
The smartest programs mix methods along the network-level versus project-level split. Screen a large network fast with a windshield pass or a specialized vehicle to find the corridors that need attention, then verify and quantify those sections with UAV imagery scored by HALO AI. The vehicle tells you where; the imagery plus AI tells you exactly what, how much, and how it is changing over time.
HALO AI is vendor-agnostic: it works with imagery from any survey-grade drone, and it can also take other sources such as survey-vehicle frames on request, so you do not have to throw out the tools that already work for you. See the input requirements on the HALO AI page.
FAQ
Method comparison questions.
ASTM D6433 defines the Pavement Condition Index method: how distress types are identified, measured, and combined into a score. It does not mandate a specific capture method. What matters is that distresses are correctly identified and measured to the standard's definitions. Drone imagery analyzed by HALO AI is scored to ASTM D6433 (and D5340 for airfields) with the deduct math shown in the report, and our optional Engineer-Verified QA add-on adds civil-engineering sign-off where a project requires it.
Yes, on request. Drone capture is the standard workflow, but if your survey vehicle or another source produces imagery that meets our resolution and georeferencing requirements, HALO AI can analyze it. Send us a sample and we will confirm whether it qualifies, at no charge. See the data requirements.
For a rough first screen, a windshield or dashcam survey is cheapest per kilometer. For a measured survey, Fadron's published comparison puts drone-based work at $90 to $120 per km against $200 to $300 per km for traditional survey, and $2,100 against $6,600 for a 3,000 m runway. But cheapest per pass is not the same as cheapest per decision: a coarse screen still leaves you without measured quantities, so weigh the cost against what the output lets you act on.
Not for everything. Specialized vehicles remain the best tool for rutting and IRI roughness at highway speed. HALO AI is strongest on objective, repeatable surface distress scoring at scale, and the two combine well: screen with the vehicle, verify and quantify with imagery plus AI.
When you want to see what the imagery-plus-AI column actually produces, see the sample survey.
Choose the method with your eyes open.
If drone plus AI is the right tool for your network, prove it on one real site: a fixed-price pilot, delivered to spec or you don't pay.