Software & Algorithms
Machine learning for machine vision
Learning methods use test data, and with that data and those images they extend, assess and classify the existing system. That makes our systems extremely flexible, so that our software can solve almost any complex machine vision task.
Does learning vision suit your task?
Send us images or samples. You will get an honest assessment, free of charge.
Anfrage senden
Machine learning at MathTech
- We use machine learning to build and optimise inspection software.
- The result is a flexible system that adapts to changing conditions while maintaining a high recognition accuracy.
- An objective validation and optimisation of the system's inspection process capability is possible as well, for standardised procedures.
Advantages
For the operator
- The user is guided when creating reference data (marking defects and labelling)
- Results are displayed and exported
- Thanks to a graphical interface the system can be used without a machine vision background
For inspection quality
- Small differences and small defects are detected and classified reliably
- Individual solutions can be built efficiently
- Fast analysis of large volumes of data, for example images
Applications
- Text recognition, including handwriting, on complex surfaces
- Complex inspection tasks with high quality requirements
- Automatic classification of objects or defects
- Detection of anomalies
- Straightforward adaptation of the inspection system to new faults, new products and changed production conditions
- Prediction of object behaviour, for example whether a solder joint will pass quality testing
- Discovery of dependencies between object features
- Machine learning
A rule set instead of a black box, where that makes more sense
Compared with deep learning approaches, an adjustable rule set has the advantage where only a very small share of production is defective, or where the defects are too subtle to be told apart from the rest without expert knowledge. It also remains traceable which rules led to a part being rejected. For individual advice on which approach suits your problem best, get in touch.
Free and without obligation
Send us your inspection task. We will tell you whether it can be solved.
A specification, a defect catalogue, samples or images are enough. In a first free feasibility assessment we evaluate the task, settle the open questions and plan the next steps with you.
Consulting, training and support are part of the deal.