01
The difficult problem
Production panels required reliable, sub-second defect inspection without requiring a historical training dataset.
← All systemsVision & Edge AI
Manufacturing computer vision
PV-Inspect
An inline visual inspection system for appliance control panels using reference-based computer vision.
VisionEdgeInfrastructureResearch
The brief
Built around the hard part.
02
The intelligence built
Industrial capture, feature matching, homography alignment and structural differencing compare every unit to a golden template.
System architecture
How the system works.
The model is one layer. The value comes from connecting inputs, intelligence and production action as one accountable system.
- 01
Triggered camera capture
Acquire and structure the operating signal.
- 02
SIFT and ORB keypoints
Transform it through the model, rules and control layer.
- 03
Homography and SSIM inspection
Deliver a decision, artifact or action into production.
In production
Where it runs.
Deployment
Industrial GigE camera, controlled lighting, Jetson edge compute and conveyor PLC synchronization.
What it changes
A zero-training-data path to automated inline inspection.
Production stack
- Python
- OpenCV
- CUDA
- NVIDIA Jetson
- GigE Vision
- PyQt