Optical character recognition on cast
Read challenging codes on cast motor parts using deep learning-based OCR
Graphical programming environment for deep learning-based industrial image analysis
An electric vehicle (EV) motor comes in a die cast housing, typically made of aluminum. Each housing is marked with an identification code, which is generally in raised letters and numbers that are part of the cast housing itself. This lettering, being of the same material as the housing, is of extremely low contrast.
This identification code is essential for traceability at the assembly facility and throughout the supply chain. It is the number to which all other records are tied. Identifying it accurately is very important since any ambiguity stops the line and requires the code be verified by human inspectors. If this happens multiple times per shift it causes significant delays and increases expense.
Traditional rule-based OCR tools typically deliver read rates up to 99 percent. For some applications that is adequate, but important applications, such as this one, require read rates to be as close to 100 percent as possible. Any human intervention to manually override the results from a failed OCR read can reduce throughput and lower efficiency.
Cognex Deep Learning’s text and character reading functions reliably and accurately deciphers deformed, skewed, damaged, or low-contrast codes. It is trained with an image sets of OCR codes with different angles, lighting, damage, and other variations.