Neural and associative modules in a hybrid dynamic system for visual industrial quality control

The development and application of neural and associative modules in the context of a hybrid and dynamic system concept for visual object inspection in industrial quality control are described. This system incorporates image processing techniques, knowledge base, and neural as well as non-neural classification methods. The system assumes a configuration based on a priori knowledge and on the results of the self-monitoring process. The first experiments and results utilizing an implemented subset of this concept are presented, with emphasis on neural and associative modules and neural hierarchies. A correspondence of neural associative memories and a conventional classification system are found. Data acquisition techniques and issues of dedicated hardware implementation are covered