Evolving task specific algorithms for machine vision applications

Increased use of machine vision system's are making a significant contribution to ensuring competitiveness in modern manufacturing. The development of task specific machine vision algorithms is a difficult process as there is no definitive model of the area so no generic approach to problem solving exists. Traditional approaches focused on the use of rule based systems to automate the generation of algorithms. However this type of approach suffers from issues related to the knowledge acquisition bottleneck and modeling of expertise. One possible solution to this problem is to evolve task specific algorithms using evolutionary tools. This work focuses on the use of an intelligent design tool that aids an engineer in designing machine vision algorithms using a hybrid intelligent system approach based around an evolutionary algorithm (EA), case based reasoning (CBR) and rule based reasoning (RBR) architectures.