In many industries, deep learning expands the possibilities for automating industrial production and can solve numerous new applications that could not previously be automated. This includes reading demanding texts, checking surface defects, checking various assembly steps, classifying product quality and sorting variable parts.

In contrast to conventional image processing, which can be used excellently for products with the same appearance, deep learning is a new approach to industrial image processing. Deep learning uses neural networks for reading markings or texts, for error analysis and to localize and classify objects. For this purpose, images of flawless and, if necessary or desired, defective products are presented to the system for learning. Using examples, a neural network learns what a product should look like and can recognize the difference between a good and a defective part in the testing process, taking expected deviations into account.

Reading of difficult industrial OCR applications

Deep learning uses artificial intelligence and uses optical character recognition (OCR) to recognize, for example, badly deformed, crooked and poorly laser-etched characters on electronic components, needle-stamped characters on metal parts, embossed characters on injection-molded products, poorly legible texts on packaging and products or characters with little contrast that conventional OCR tools cannot decode.

Quick and precise assembly checks

The artificial intelligence enables a reliable automated assembly check because even complex features and objects can be recognized. The system can be trained to build a comprehensive library of components that can be located in the image, even if they are visible at different angles or of different sizes.

Identifies unpredictable errors

The deep learning system is suitable for finding anomalies on complex parts and surfaces, even in situations where the appearance of the defects can be unpredictable. The system is trained with images of faultless parts in order to identify image areas in the inspection process that deviate from the normal appearance, whereby defective parts can be recognized.

Get in contact
This site uses cookies. Further information can be found in the privacy statement.