Network-wide road defect audit in days, not months, at TMH9 standard.

The authority was required to produce an updated defect condition report for a 180+ kilometre provincial road network as input to the following financial year maintenance budget. Previous audits had been conducted on foot by rotating inspection teams, a process that took four to six months and returned subjective, inconsistent grading. Budget deliberations were moving faster than the audit cycle could keep up.

The authority needed three changes. Objective and repeatable defect classification. GPS-accurate defect positioning so maintenance tenders could be issued against verified locations. And full network coverage in a timeframe that fitted the budget decision window. TMH9 compliance was non-negotiable.

Off-the-shelf pavement management systems either required dedicated survey vehicles the authority did not own, or returned output that could not be aligned with TMH9 without significant manual work.

Delta Scan had been developing RoadAI, our proprietary AI road defect detection pipeline, through a series of smaller client engagements. This project required it at network scale for the first time. We proposed a three-phase engagement using a vehicle-mounted capture rig and our in-house inference pipeline.

Phase one was scoping. We agreed the defect taxonomy to TMH9 classes: potholes, cracks (longitudinal, transverse, alligator), ravelling, edge break, rutting, and patching. Distance-based frame sampling parameters were set to ensure no double-counting and no missed defects at highway speeds. The capture vehicle was rigged with a Sony camera array, RTK GPS, and wheel-speed distance reference.

Phase two was capture. Over three operational days, the vehicle traversed the full network at normal traffic speeds, logging 34,309 georeferenced frames to DeltaCloud. Live quality monitoring flagged out-of-spec frames for re-capture on the same day, eliminating the need for return visits.

Phase three was inference and reporting. RoadAI processed the full dataset in days rather than the weeks a manual review would have taken. Each detected defect was logged with GPS position, severity classification, photographic reference, and TMH9 class code. The final deliverable was a full audit report with a map-based dashboard hosted on DeltaCloud, plus a CSV export for import into the authority existing asset management system.

• Capture vehicle. Delta Scan mobile survey rig
• Cameras. Sony Alpha 4K array (forward and lateral)
• Positioning. RTK GPS with wheel distance sensor
• Inference software. RoadAI (Delta Scan proprietary)
• Delivery environment. DeltaCloud with map-based dashboard

• Day 0. Scoping, taxonomy alignment, rig preparation
• Day 1 to 3. Vehicle-mounted capture across 180+ km network
• Day 4 to 7. RoadAI inference, engineering review, dashboard build
• Day 8. Dashboard and CSV export delivered to authority

The authority used the dashboard to prioritise the following year road maintenance budget by severity and region, moving from a subjective team-graded process to an objective, TMH9-aligned audit.

The dashboard has been retained as the running condition record for the network, with quarterly re-capture now scheduled to maintain an up-to-date picture. Delta Scan has been engaged for the next cycle of captures.