Home / Cities and public spaces / Pothole detection

Pothole detection AI that gets road damage repaired inside the contract deadline.

Potholes are often reported by drivers after the damage is done. ZoneIntel AI runs pothole detection AI on your street cameras. Each day it compares a marked stretch of road with its last clear image, and a reviewer confirms each change.

Each defect has a repair deadline

Each type of damage maps to a repair class and a deadline in the road maintenance contract. A missed deadline leaves a hazard on a busy road.

DamageRepair classTypical deadline set by the city
Pothole in a running laneUrgent24 hours
Pothole on a cycle route or crossingUrgent24 hours
Missing or broken drain coverUrgentSame day
Broken road edge or kerbPlanned7 days
Cracking and worn surfacePlannedNext resurfacing list
Faded lane markingsPlannedNext marking round

Results

ResultPilot measure
Damage found before a complaintDefects found on camera, against resident reports
Repairs on timeRepair orders closed inside the deadline, per contractor
Fewer inspection drivesInspector trips per week, before and after

Each damaged stretch has a repair order, a deadline and a photo of the patch. The contractor report shows which repairs ran late.

Road damage in lane06:30
Street camera view with a damaged section of road marked and a motorbike riding through it

The marked section of the nearside lane at camera S-31 has changed since the last clear image, with water standing in a damaged patch.

Where
Street camera S-31, market road
Sent to
Road maintenance contractor duty officer, by work order and Teams
If nobody responds
Public works is alerted at the 24-hour repair deadline, then the city operations centre.
Example alert with made-up names. Simulation.

How a repair order closes

The repair order stays open until the reviewer sees the surface patched on the same camera view. The times go into the contractor report.

Street camera with the damaged section of the nearside lane marked
06:30, found
Repair order RO-2207Street camera S-31, market road
  1. 06:30Damaged section found in the nearside lane at camera S-31.
  2. 06:31Repair order to the road contractor, 24-hour deadline.
  3. Day 1, 16:00Crew on site, seen on camera; lane coned off.
  4. Day 1, 19:20Patch confirmed by the reviewer on the same camera view, order closed.

Questions about pothole detection AI

How does pothole detection AI work?

Pothole detection AI compares images of the road surface over time to find damage. ZoneIntel marks a watch area over each stretch of road and checks it daily. A reviewer looks at each change and raises the repair order.

Can fixed cameras see road damage?

Fixed cameras see the stretch of road in their view. That covers junctions, bus stops and busy streets, where damage matters most. Cameras on vehicles can be added to survey the whole network.

How does ZoneIntel AI rank repairs?

The reviewer sets the repair class from lane position and traffic. A pothole in a running lane outranks a crack at the kerb. The city sets the deadline for each class.

Does it check that the repair was done?

ZoneIntel checks the same view after the repair deadline. The order closes when the reviewer sees the patch. A repeat failure at the same spot is flagged in the contractor report.

Does it work in rain?

Standing water can hide a pothole, so a wet watch area is checked again once the road is dry. Water that stays after rain is raised as a drainage case.

Road check on your own cameras

Share a day of footage from a few street cameras through a secure link. We send back the damage ZoneIntel AI would have raised, with photos.