Professional Case Study

Industrial surface-defect inspection

Detecting small defects in coil images

Project context

Professional • Aperam Genk

How the inspection works

Open video

Two cameras inspect the top and bottom surfaces separately. For example, the top camera may see a scratch while the bottom camera sees a dent. Each defect is classified by type and severity from 1 to 3. Severity 3 on either surface means scrap the strip. Generic illustration: the equipment and defects are fictional, not an Aperam installation or an inspection result.

Problem

Small surface defects can be hard to find in high-resolution images of coils. The task was to detect defects at the millimetre scale and make inspection more reliable.

My contribution

As a Computer Vision Engineer at Aperam Genk, I built a Python algorithm using OpenCV and PIL to detect these defects. This was part of my work on open-source algorithms for inspection and quality assessment.

I also mentored junior engineers and worked with the innovation team to identify other uses for computer vision.

Outcome

The inspection solution was rolled out across multiple plants and improved inspection reliability.

Professional context

I worked on this during my Computer Vision Engineer role at Aperam Genk, from September 2019 to June 2024. My earlier work on predictive maintenance with sensor networks was a separate project.

← Back to selected work Contact Alexandre