Machine Vision and Optical Character Recognition (OCR)

Definitional Boundaries & Taxonomy

EN: Machine Vision (MV) in the context of industrial automation refers to the engineering discipline of extracting actionable information from digital images to control manufacturing processes, execute robotic trajectories, and perform automated inspection. Unlike computer vision, which is a broad academic field focused on image understanding, Machine Vision is strictly deterministic, prioritizing bounded execution times (real-time constraints), high reliability under harsh conditions, and seamless integration with Programmable Logic Controllers (PLCs) or robot kinematics. Optical Character Recognition (OCR), traditionally a subset of document analysis, has evolved within MV into Industrial OCR (often termed Optical Character Verification - OCV). Industrial OCR entails reading alphanumeric codes, DataMatrix, and serial numbers directly marked on physical substrates (Direct Part Marking - DPM) such as metal, glass, or rubber, under extreme variations in lighting, surface topology, and manufacturing tolerances. The taxonomy of MV splits into: 1) 1D/2D Barcode/OCR verification, 2) 2D Metrology and defect detection, 3) 3D Vision for robotic bin picking and volumetric analysis, and 4) Deep Learning-based visual inspection for subjective or highly variable anomalies.