Axle configuration classification with computer vision

Whether a vehicle is overloaded depends on its configuration: the number of axles, how they are grouped and the vehicle type. Selecting this by hand at busy stations is slow and error-prone.

What the model does

Side-view cameras capture each vehicle. A computer vision model detects wheels and axles, groups them and classifies the configuration. The software then applies the corresponding limits.

Combine with sensor data

Weigh-in-motion sensors and loops already measure axle count and spacing. Using camera and sensor data together gives a more reliable classification than either alone.

Handle the hard cases

• Lifted axles on trucks and trailers.
• Unusual or modified vehicles.
• Vehicles partly hidden by others.

Uncertain classifications should go to an operator to confirm, and those confirmations become training data.

Measure accuracy properly

Test on images from the actual sites, across day, night and weather, and report accuracy per vehicle class — not just one overall figure.

See our AI vehicle classification work.