Cognitive image processing focuses on understanding images and image sequences. This requires adaptive and trainable components capable of generalising and interpreting within a certain context. Research mainly focuses on the interpretation of activities in production environments such as by automatic, semiautomatic or manual production processes to acquire high-level skills and to detect deviations from “normal” processes. The approach is to build upon existing machine learning methods and to adapt them to the specific requirements of machine vision systems.

Examples include learning of complex decision-making based on visual input, acquiring skills that require 2D or 3D visual feedback and detecting abnormal activities in semi-automatic production processes.