Deep learning has hugely expanded what machine vision can inspect. On lines where defects are variable or hard to describe with fixed rules —like canning lines— it’s the technology that makes the difference. Let’s look at how deep learning is applied to quality control and to advanced can inspection.
Deep learning applied to quality control
Unlike rule-based vision, deep learning learns from examples of good and defective parts. This way it recognises defects that can’t be defined by a formula: texture variations, subtle deformations, irregular defects. It’s ideal when product variability is high.
The challenges of inspecting cans
Canning lines are demanding: high speed, metallic surfaces with reflections, and defects such as dents, seam/seaming defects, poor sealing, deformations or illegible coding. Many of these defects are variable and hard to parameterise.
Advanced can inspection with deep learning
With models trained on real line images, the system learns to tell a correct can from a defective one even with reflections and variability. It detects dents and seam defects, verifies correct sealing and validates coding, automatically rejecting non-conforming units.
Benefits
100 % inspection on high-throughput lines, detection of defects that fixed rules miss, fewer complaints and traceability of every reject.
At AIS Vision Systems we apply deep learning and our inspection platform to advanced can and canning inspection. Tell us about your line.