FOR INDUSTRIAL COMPUTERTOMOGRAPHY
How we use artificial intelligence
Microvista is considered a pioneer in the field of automated analysis and evaluation routines of CT scan data in the industrial sector. Building on our many years of experience, we have advanced the use of AI processes in our software solutions and offer you individual AI solutions for industrial use with a high degree of robustness against fluctuations in the measurement process.
Microvista offers customised IO/NIO assessments and/or quantification of discontinuities using robust AI techniques.
- Object classification according to IO/ NIO
- Object detection and localisation of e.g. fractures, chip residues, pores
- Object segmentation for 2D and 3D pore analysis: detection, localisation and quantification (area, volume, diameter)
By combining AI methods with classical approaches, we are able to perform pixel-precise measurements of geometry features.
Our service extends along the entire development chain, starting from the problem definition to the implementation of the ML solution in existing systems. This includes the generation of training data, the selection and training of a suitable ML architecture and its combination with classical image processing approaches as well as the validation of the ML solution.
This is how we do it
1. Solution approach: In a first conversation we ask you for an exact description of the problem, this can be e.g. a boundary pattern catalogue. Based on this information, we can quickly assess whether our AI approaches are suitable for your concern. We then discuss how we can set up and train our AI for your use case as well as implement it in your existing system.
2. Learning data: If you already have component scans available, you can simply provide them to us. If no scans are available, it is best to send us sample parts from which we can create suitable scans. The learning and test basis for the AI solution is built from the data material. If you do not have enough data, it is part of our expertise to deal with this problem and still enable the training of an AI solution with the help of synthetic data sets.
3. Training: The AI is then trained, i.e. the evaluation of your images is learned, and what is learned is subsequently validated together using reference data. The training is an iterative process that aims to achieve sensitivity and specificity of the AI solution according to the testing requirements.
4. Implementation: The developed AI solution is then supplemented by further software components (e.g. result processing, further analysis tools) and implemented directly into your existing system. We also offer you integration into our own software environment, which takes care of everything from recording the component number and other master data to creating a customised test report.