Conveyor defect inspection
Deep-learning inspection that flags defective parts on a moving line.
Development stage
Stage 3 of 8: system architecture.
- Problem identification
- Research
- System architecture
- Simulation
- Prototype
- Testing
- Validation
- Production
Problem
Visual quality checks on production lines are subjective and suffer from fatigue.
Solution
A camera over the line captures each part; a trained model classifies defects and signals the line to reject.
Technology
- Computer vision
- Deep learning
- Industrial automation
Hardware
- Industrial camera
- Lighting rig
- Edge computer
- Reject actuator
Software
- Python
- Training pipeline
- PLC interface
Target outcomes
- Consistent, documented inspection decisions
- Defect data for process improvement
- A path to in-line deployment
Team and media
Developed by the VEL Robotics R&D team.
Photos and video of this project will appear here.
Tell us what you want to build, teach or research.
Pick the kind of enquiry and it goes to the team that handles it.