AI Road Damage Detection
Road damage detection uses a camera fitted to a municipal vehicle to flag potholes, cracks, sunken kerbs and worn road markings during the vehicle's normal duty round. No separate inspection vehicle is sent out: a refuse truck, a municipal enforcement vehicle or a service minibus has already scanned the route it drives anyway. Each detection, from a single pothole to a broken kerb, is recorded with its location, date and photo and assigned to the public works team as a task. Before the crew goes out, it already knows what is wrong on which street.
What is road damage detection?
Road damage detection is a field inspection method that picks out road surface deterioration from images taken from a moving vehicle. Potholes, surface cracks, asphalt subsidence, broken kerbs, damaged manhole covers and faded road markings are flagged as separate classes, so pothole detection is one part of a wider road surface inventory.
The traditional methods are citizen reports and periodic site rounds, and both leave coverage narrow. A report covers only the spot someone complained about; a site round covers only the route driven that day. Vehicle-mounted scanning turns the routine routes of service vehicles into an inventory without any extra trips.
The output is not a flat list but an inventory positioned on a map. When the same street is scanned on different dates, you can compare whether the damage is getting worse, and maintenance priority is set from that comparison. A pothole that has grown since the last pass moves up the list; damage that has not changed can wait for the scheduled programme.
Because scanning rides on vehicles that are already on the road, coverage grows with the routes your fleet drives rather than with the number of dedicated inspection rounds. Streets that nobody reports and no inspector visits still end up in the inventory.
How does road damage detection work?
Scanning follows three steps.
A camera is fitted to the vehicle. A camera that sees the road and an Algi Edge device that receives location data are installed in the vehicle. It does not have to be a dedicated inspection vehicle; any service vehicle with a routine route is suitable.
Video is processed in the vehicle. Analysis runs on the on-vehicle Algi Edge device while driving. What leaves the vehicle is not raw video but a detection record: damage type, location, date and a single photo. This keeps the load on the data connection low. Where there is no mobile coverage, records are stored on the device and transferred once the connection returns.
Detections become tasks. Records are collected on a map and filtered by neighbourhood, street and damage type. Selected detections are assigned to the public works team as work orders and closed when the repair is done. When the same route is scanned again, you can see whether a repaired spot has really been fixed.
The inventory also shows which areas the year's maintenance budget is going to, so spending can be checked against the actual condition of the roads.
Which industries use it?
Frequently asked questions
- What is road damage detection and how does it work?
- Road damage detection evaluates footage from a camera fitted to a service vehicle while it drives. Potholes, cracks, subsidence, broken kerbs and worn road markings are flagged as separate classes. Each detection is recorded on a map with its location, date and photo, and assigned to the public works team as a task.
- Does road damage detection need extra hardware?
- Each vehicle needs a camera and an Algi Edge device that receives location data; beyond that, there is no infrastructure investment. You do not need a dedicated inspection vehicle. Vehicles with routine routes, such as refuse trucks, municipal enforcement vehicles or service minibuses, are enough for the job.
- Are my existing IP cameras compatible?
- This solution does not work with fixed IP cameras but with a camera fitted to the vehicle, because the road surface can only be scanned while moving and from a close angle. The municipality's existing fixed city cameras are used for junction, square and bus stop analytics, not for the road surface inventory. The vehicle camera is supplied with the installation.
- How long does setup take, and when does it go live?
- Mounting the camera and device in a vehicle takes a few hours, and the first vehicle can start scanning the same day. Rolling out to the fleet depends on the number of vehicles. For a meaningful inventory, the routes need to be scanned once from start to finish; depending on the size of the district, this takes a few weeks.
- Is road damage detection GDPR and KVKK compliant?
- The system looks at the road surface; it does not recognize or identify pedestrians or drivers. Data such as faces and number plates does not enter the detection record. Video is processed on the Algi Edge device in the vehicle, and only the damage type, location and a single photo are transferred. Records follow the municipality's existing data retention policy.
Related rules from the library
ALL →OBJECT_DETECTION
Damaged-Road Detection
Vehicle-mounted cameras detect potholes and defects; each finding becomes a geolocated work order.
OBJECT_DETECTION
Asphalt & Pavement Deformation
Deformations on asphalt and pavements reported automatically along the route.
OBJECT_DETECTION
Road-Marking Wear Analysis
Detects visible wear on lane markings; feeds the renewal plan.
DON'T SEE YOUR NEED?
Dijital Atölye listens to the problem in the field, prototypes, validates and productizes. Some of the products running in the field today were born on this line.
Bring the problem, let's talk prototype →