A classification count is only as good as the sensor underneath it. Five technologies are in common use for FHWA 13-class counts, and each one fails in a different place. This comparison lays out what each can and cannot do, so the choice matches the study instead of the habit.
If you need the background on what the 13 classes are and why they matter, start with our explainer.
The FHWA scheme is defined by axles and units. So the real question for any sensor is simple: can it count axles, or is it guessing the class from something else? Length-based systems can separate cars from trucks reliably, which is enough for a three- or four-bin summary. They cannot reliably separate a Class 5 from a Class 8, or a Class 9 from a Class 10, because those overlap on length. Full 13-class accuracy needs axle information or a way to see the vehicle's configuration.
How it works: Two rubber tubes across the lane. Each axle sends a pressure pulse, and the spacing between pulses gives axle count and spacing, which maps to a class.
Strengths: Cheap, portable, genuinely axle-based. The default for short-duration counts for decades.
Where it falls short: Tubes get cut, kicked loose, and flattened by heavy trucks. Multi-lane sites and heavy congestion produce overlapping pulses. A crew has to set them and come back to pull them, which puts people on the road twice. Rain and snow degrade the data. On high-volume highways, tube counts are often the first thing a reviewer questions.
How it works: Loops cut into the pavement detect presence and length. Piezo strips detect each axle. Together they give a full axle-based classification.
Strengths: The permanent-station standard. Accurate when calibrated, and the basis for most continuous count sites.
Where it falls short: It requires a pavement cut, lane closures to install and repair, and ongoing calibration. Loops miss motorcycles. Pavement movement and resurfacing break them. The cost per site makes them impractical for the dozens of short-duration locations a study program needs.
How it works: A camera with vision software counts and classifies vehicles from the image.
Strengths: Non-intrusive, easy to mount, and provides a visual record. Good classification in daylight with clear sightlines.
Where it falls short: The image is the only input, so when the image degrades, so does the count. Rain, fog, glare at sunrise and sunset, and darkness all cost accuracy. Axle counting from video alone is unreliable at night and in weather, which pushes the system back to length-based guessing on exactly the days the study cannot afford it. Our post on AI versus manual monitoring covers where vision helps and where it needs backup.
How it works: Side-fire or forward-fire radar measures speed, range, and length of each vehicle.
Strengths: Works in every weather condition and at night. Very accurate on speed and volume. No pavement work.
Where it falls short: Radar sees length, not axles, so it classifies into length bins. That is fine for car-versus-truck but not for the full 13-class scheme. Occlusion on multi-lane roads, where a truck hides a car, is a known weakness.
How it works: One device combines HD3D radar with AI video and fuses both on the sensor. Radar supplies speed, range, length, and presence in any weather. Video supplies axle count, trailer gaps, and body type. The fusion resolves what neither can alone.
Strengths: Full 13-class output without pavement work, in rain and darkness, with a video record for QA. One install, no return trip, and the same device can run intersection, work zone, and parking applications. See our breakdown of HD3D radar versus traditional sensors for the radar side in detail.
Where it falls short: It costs more per unit than a set of tubes. It needs a mount with a view of the lanes, which is usually a pole, sign structure, or trailer. And like any sensor, it needs a short calibration at setup.
For a continuous count station that is already cut into the pavement and calibrated, loops and piezos keep working. For a short-duration classification program with many sites, the trade is between the low unit cost of tubes and the redo rate they carry. One lost 48-hour count, with a crew sent back out, erases the savings on several sets of tubes.
That is the gap FusionSensor was built for: 13-class counts at 98 percent or better, in any weather, from one device that goes up once. If you are specifying a program, our classification count checklist walks through site selection, duration, and QA, and the traffic data collection page covers outputs and integrations.
Subscribe to The Edge Brief for monthly field notes, deployment stories, and product updates.
Subscribe →