Designing a defect classification system: A case study

摘要

This paper describes designing and testing a vision system for web material inspection. The system incorporates five subsystems: sensing, detection, characterization, feature analysis and classification. Each subsystem contains adaptive image processing and pattern recognition algorithms that perform specific functions. A very difficult real-world case study of defect classification is detailed and examples of algorithms used by the five subsystems are outlined. The results are summarized and interpreted.