Flexible and adaptive machine vision systems

Machine vision systems cover a wide range of processing steps from low-level processing to high-level decision-making and cognitive systems. Quite often the single steps have to be narrowly tuned to a specific application, which limits their applicability and flexibility. Profactor’s research activities aim at creating adaptive systems that are able to learn from human behaviour and to adapt to changes in the environment. The focus is not only on the higher level decision making, but also on low-level processing steps. We use and develop image processing methods for texture-based segmentation, video interpretation, and high-level decision making. We use machine learning and classification methods such as (recurrent) neural networks, decision trees and rule-based methods. A particular emphasis is on including the “human element” in the adaptation process. 

Thermography

Thermography is of particular relevance for industrial applications and enables us to see defects hidden under the surface. It extends our capabilities beyond the visual domain. Research focuses on optimizing the setup of thermography systems for in-line inspection in order to maximize contrast and to detect even the smallest cracks.