Visual Quality Control in Manufacturing
Computer vision system for automated defect detection on production lines. 99.2% accuracy, 200 parts/min throughput.
Challenge
A major manufacturing conglomerate relied on manual visual inspection on its conveyor lines. Operators checked each part for defects — chips, cracks, geometry deviations. At 200 parts per minute, the human eye inevitably missed up to 15% of defective items. Fatigue and subjective assessment led to customer complaints and financial losses. The client needed an automated system operating in real-time without slowing down the production line.
Solution
A computer vision system based on convolutional neural networks was developed and integrated directly into the production line. High-resolution industrial cameras capture each part from multiple angles. A real-time detection model classifies defects by type: cracks, chips, scratches, geometry deviations, foreign inclusions. When a defect is detected, the system automatically rejects the part and generates a report with photo evidence. The model is continuously retrained on new defect types without stopping the conveyor.
Results
Technologies
Approach
Production line audit and defect type analysis
Examining the production process, cataloging defect types, and defining detection criteria.
Dataset collection and labeling
Collecting and annotating over 10,000 images of parts with various defect types.
Detection and classification model training
Training convolutional neural networks for real-time defect detection and classification.
Camera and compute module integration into the conveyor
Installing industrial cameras and edge computing modules directly on the production line.
Similar challenge?
Tell us about your project — we will propose the optimal solution.
Discuss a project