Every FHWA 13-class count runs into the same problem eventually. A truck passes at a head-on angle, another vehicle blocks the axles, or rain cuts visibility, and the system has to either guess a class it cannot back up or report the vehicle as unclassified. Either way, the gap ends up in your report.
GoodVision runs the complete FHWA 13-class scheme, the U.S. national standard for vehicle counting and classification, and it is built around that exact failure point.
The scheme spans light vehicles through the largest combination trucks:
Light vehicles: 1 Motorcycles, 2 Passenger cars (including SUVs), 3 Pickups, vans and other light 2-axle vehicles, 4 Buses
Single-unit trucks: 5 Single-unit, 2 axles, 6 tires (box trucks), 6 Single-unit, 3 axles, 7 Single-unit, 4 or more axles
Combination trucks: 8 Tractor-trailer, up to 4 axles, 9 Tractor-trailer, 5 axles (the everyday semi), 10 Tractor-trailer, 6 or more axles, 11 Multi-trailer, up to 5 axles, 12 Multi-trailer, 6 axles, 13 Multi-trailer, 7 or more axles
Class 14 covers anything that does not fit the scheme, such as oversize loads or machinery. Multi-trailer classes are where most camera-based systems fall short. GoodVision covers the full range.
For every vehicle that passes, GoodVision reads three things from the footage: what it is (body shape separates a motorcycle from a car, a bus, or a truck), what it is pulling (tracked across the scene to identify a single unit, one trailer, or two or more), and how many axles it has, when the wheels are in view.
Most classes never depend on that last step. Motorcycles, passenger cars, light 2-axle vehicles, and buses are identified from body shape alone, so classes 1 to 4 return exact every time, whether or not the axles are visible.
Trucks are different. When an axle count is available, a truck resolves to its exact FHWA class, 5 through 13. When it is not, because of a head-on angle, an occluded view, or poor conditions, GoodVision still knows the body type and the trailer count, so it reports the correct group instead of a class it cannot stand behind:
No wrong class, and no unclassified vehicle. Groups and exact classes add up together, so totals stay consistent across a report.
The classes themselves never change, which keeps reports consistent across sites and over time, but how often you get the exact class depends on what the camera can see.
A single overview camera is the standard setup. Every vehicle gets classified, cars, buses, and motorcycles at their exact class, and trucks mostly at the group level.
Add a roadside axle camera, and most passing trucks resolve to their exact class instead. A truck the system cannot fully see still falls back to the correct group automatically, so there is never a hole in the data.
Most video analytics systems handle FHWA classes 1 through 4 well enough. Body shape is a solved problem. Trucks are where it breaks down, because the standard scheme depends on axle counts, and a roadside camera does not always get a clean view of every axle on every truck that passes.
The common fallback is a single, fixed camera angle and a system that reports whatever class the algorithm is most confident in, correct or not. That produces a count that looks complete but is not verifiable. If an engineer later needs to defend those numbers to a client or a regulator, "the model was fairly confident" is not an answer that holds up.
The alternative some vendors use is to drop unclear vehicles into an "unclassified" bucket. That is more honest, but it just moves the gap. A corridor with a meaningful share of unclassified heavy vehicles is a corridor whose truck-class breakdown you cannot fully trust.
FHWA 13-class classification runs inside GoodVision Video Insights (GVVI), the on-demand traffic study product. The workflow stays the same as any other GVVI study: upload recorded footage from a standard traffic camera or drone, and the AI processes it and returns results within hours. No new hardware, and no manual review pass to fix misclassified trucks.
For sites where axle visibility varies by lane or by weather, that is the difference between a report you have to caveat and one you can hand over as is. A study that mixes a clear side-angle segment with a partial, head-on segment still comes back internally consistent: the vehicles resolve to exact classes where the axle count supports it, and to the correct group everywhere else.
GoodVision ran a vehicle-classification survey on live U.S. interstate traffic, including sections of I-35 and I-94: real corridors, real mixed flow, from passenger cars to five-axle semis and multi-trailer combinations. The classifier handled the full FHWA range on the same kind of camera feeds most agencies already run.
Classification reads appearance only. There is no number-plate reading, no vehicle identity, and no tracking of individuals, which keeps the data usable for planning and reporting without raising privacy concerns on public corridors.
That combination, full FHWA 13-class coverage plus a defensible fallback for the vehicles a camera cannot fully see, is what makes counts usable for FHWA-standard reporting rather than an approximation you have to explain away.
Book a demo at goodvisionlive.com/request-demo/ and see FHWA 13-class classification running on your own highway camera feeds.