Machine Vision–Based Quality Inspection
Problems of manual quality inspection
Sample inspection
– Possible only in cases with a high probability of occurrence
– Delayed recognition of quality changes due to changes in manufacturing process conditions
Full inspection
– Labor cost
– Months of training required for the person in charge
– Frequent employee turnover occurs
– Inspection speed affects product production capacity
– Different results depending on various factors such as work time, day of the week, and individual differences
Visual Insights
– By applying deep learning–based analysis technology, higher accuracy can be achieved
– Possible to apply analysis for new products and scenarios with several days of training
– Easy to apply and expand – can be applied to any type of image
– Costs can be reduced by using general-purpose hardware
– No downtime is required for upgrades using Cloud technology
– A big-data analysis platform for not only fragmentary inspection but also optimization scenarios across multiple work units
PCB-Vision
– Quality defect vision inspection solution
– Practical application possible with simple additional data training based on learned algorithms
– Compatible with existing inspection equipment and applications
– Detects defects that may be missed by visual inspection with close to 99% accuracy
Application Areas
– PCB, FPCB quality (vision) inspection
– Deep learning solutions can be applied in various fields where machine vision is used
– More than 10% improvement in defect detection rate compared to existing machine vision (varies depending on application area)
