Road and signage condition from cameras: what can be automated, and how to test it

A fixed camera sees the same scene all day. A condition that stays visible for hours, such as a fallen sign or an object in a lane, is therefore a better candidate than a small detail that needs a close look. Cameras are weakest at small surface defects, because they look at the road from a distance and at a shallow angle.

1. Capture. Cameras that overlook the corridor are connected to SquintPRO. They can be existing cameras if their angle and resolution suit the condition you want to check.
2. Define. Each condition is set up from SquintPRO's trained detection libraries where one fits, or by describing it in text or with a reference image, without a model-retraining project.
3. Detect and record. When SquintPRO sees a suspected case, it logs the camera, time, zone and type of condition, with a video clip as evidence.
4. Review. A person confirms or dismisses each case before anything else happens, so field teams are not sent to cases nobody has checked.
5. Refer. Confirmed cases go to the maintenance team or the responsible department through the alert route or system your organisation chooses. Interkey builds that connection during implementation.
6. Measure. Cases are counted by area, period and type of condition, so you can see where the same problems recur.

They add to inspections when:

• the same stretch keeps producing the same problem and needs watching between visits;
• a condition appears outside inspection hours, for example an object left in a lane overnight;
• you need evidence of the condition before and after repair, to close the case.

They do not replace inspections when:

• the asset is outside camera coverage;
• the judgement needs something the image cannot show, such as structural condition or the depth of a defect;
• the fix needs a site visit anyway.

For municipal teams working on visual pollution, a separate Arabic guide covers the elements in the Ministry of Municipalities and Housing initiative: رصد عناصر التشوه البصري بتحليل الفيديو

Relatedرصد عناصر التشوه البصري بتحليل الفيديو

• A camera list. Where each camera is, what it overlooks, its condition, and who owns and operates it.
• A written definition of each condition. What counts as a case, what does not, and which part of the view it applies to.
• A review and referral route. Who reviews cases, who receives confirmed ones, and in which system.
• Where processing runs. SquintPRO can run on your own infrastructure, where the footage does not leave your premises, or in the cloud with 3, 6 or 12 months of retention chosen at setup.
• Privacy rules. Road cameras capture passers-by and vehicle number plates. Decide with your legal team who may view footage and cases, and how long they are kept.

Agree these measures in writing before go-live, and run the pilot over a period that covers weekdays, weekends and night hours:

• Confirmed cases as a share of all cases raised, to see how much of the review workload is useful;
• What the cameras missed: conditions found by inspection on the same corridor in the same period that the cameras did not raise;
• Time from detection to referral, and from referral to close-out;
• Daily review workload, against what your team can handle.

Then decide in writing whether to extend, adjust or stop.

SquintPRO is Interkey's own video-analytics platform. Interkey's team deploys and tunes it from Riyadh, and it works on existing CCTV. It is deployed in a smart-city project to monitor road and sign condition, alongside deployments in energy, at a port and in Interkey's own factory.

The platform includes 35 trained detection libraries across 12 core use-case categories, and a new condition can be added by describing it in text or with a reference image. It is priced by quote, according to the site, the number of cameras and the use cases in scope.