Project · Road Inspection · USA

Engin.AI: Measuring Road Damage from Camera Images in Real-World Dimensions

Your contact for a similar project is Daniel Giebelhaus in Zurich. Our development team does the implementation.

Client
Engin.AI
Location
USA
Industry
Road Inspection
Period
Early 2025

15 days

until the measurement module was delivered, including discussions and revisions

2 days

until the first working version of the data processing

Starting point

Engin.AI already had a system that detects cracks, potholes and other damage in camera images. Two things were missing: converting pixels into real-world dimensions on the road surface, and clean, structured data instead of raw segmentation files.

What we built

Measurement module: The camera is calibrated on site to the road plane using four corner points. A perspective transformation then provides the length and area of every instance of damage.

Data processing: Contours are generated from the segmentation masks, cracks are represented as polylines, and everything is exported as JSON with damage type and coordinates.

Result

The measurement module was delivered in 15 days, including discussions and revisions. A first working version of the data processing was ready after 2 days and was refined afterwards.

Technologies

PythonOpenCVscikit-imageskanNetworkXGoogle Colab

Describe your task in two sentences

We will tell you whether a dedicated tool is worth it and what it would take.

Describe your task