AI Applications · Energy and Facility Inspection
Assisted Bridge Surface Defect Inspection
Bridge inspection has two hard problems: getting close enough to the underside, and seeing it the same way every time. Aircraft solve reach; the detection step helps with consistency.
Overview
- Crack width needs a calibrated scale in frame — width cannot be inferred from pixel counts without known geometry and distance.
- GPS is unreliable under a deck; positions are tied to chainage markers or visual references that match the inspection record.
- Wet patches, staining, shadows and formwork joints are the main sources of false candidates and need to be part of the review.
- Output fields should map to the inspection standard the client reports under, so the result fits the existing record rather than creating a parallel one.
Two problems: reach, and consistency
Bridge inspection has always had difficulty reaching the interesting parts. Deck soffits, bearing shelves, pier faces and the inside of box girders either need access equipment or closing lanes, and sometimes both. An aircraft changes the cost and risk of being there.
The second problem is less visible and just as important: two inspectors looking at the same crack may record it differently, and the same inspector may record it differently from one year to the next. Repeatable imaging does not solve this on its own, but it gives every subsequent inspection the same starting material.
What the imagery has to carry
The most consequential requirement is scale. A crack width is a physical dimension, and it cannot be derived from pixel counts unless the geometry and distance are known. In practice that means a calibrated reference in frame, or a ranging measurement recorded alongside the image. Without it, the imagery is usable for locating features and unusable for measuring them.
Positioning under a deck is its own problem. GNSS is unreliable beneath a structure, so locations have to be tied to chainage markers, pier numbers, or visual references that already appear in the inspection record. If the delivered position system does not match the one the inspector uses on site, the imagery becomes difficult to correlate with anything.
Lighting and surface condition drive most false candidates. Damp patches, staining, shadow boundaries and construction joints can each look like cracking when presented as a flat image. These have to be anticipated in review rather than discovered as surprises.
Fitting the result into the existing record
An inspection report follows a template set by the standard the client reports under, and the automated step should populate that template rather than invent a parallel one. If the output does not map onto the fields an inspector already completes, it becomes another artefact to reconcile.
Configuration matters here. Dedicated bridge inspection configurations address lighting, stand-off stability and approach paths for structure work, which differs from open-terrain inspection flying.
Roofs and building surfaces are the related case — see roof and building surface inspection — where access is easy but drainage staining produces many of the same false candidates.
The limit that must stay stated
Assisted detection helps locate and record. It does not decide condition, it does not assign load bearing significance, and it does not replace the judgement of a qualified bridge inspector. Where a finding has structural consequence, that consequence is assessed by someone entitled to assess it.
Inspection is also not a single flight. It is a periodic activity whose value comes from comparison across cycles, which brings us back to repeatability: the best flight today is the one that can be flown the same way again.
Related pages
- M-15 with thermal imaging optionView
- M-30 heavy-payload long-endurance multirotorView
- AI and Enterprise System IntegrationView
- Development Testing and DeliveryView
More in this area
- Assisted Detection of Solar PV Thermal AnomaliesRead
- Screening Visual Anomalies on Transmission AssetsRead
- Surface Defect Screening on Wind Turbine BladesRead
- Thermal Anomaly Patrol for Substation FacilitiesRead
- Change Detection along Pipeline CorridorsRead
- Recording Storage Tank Appearance AnomaliesRead
- Assisted Bridge Surface Defect InspectionThis page
- Roof and Building Surface InspectionRead
- Linking Asset Registers to ImageryRead
- Defect Review Work Orders and Report LoopRead