A reference architecture for flexibly integrating machine vision within manufacturing
A reference architecture provides an overall framework that may embrace models, methodologies and mechanisms which can support the lifecycle of their target domain. The work described in this thesis makes a contribution to establishing such a generally applicable reference architecture for supporting the lifecycIe of a new generation of integrated machine vision systems. Contemporary machine vision systems consist of a complex combination of mechanical engineering, the hardware and software of an electronic processor, plus optical, sensory and lighting components. "This thesis is concerned with the structure of the software which characterises the system application. The machine vision systems which are currently used within manufacturing industry are difficult to integrate within the information systems required within modem manufacturing enterprises. They are inflexible in all but the execution of a range of similar operations, and their design and implementation is often such that they are difficult to update in the face of the required change inherent within modem manufacturing. The proposed reference architecture provides an overall framework within which a number of supporting models, design methodologies, and implementation mechanisms can combine to provide support for the rapid creation and maintenance of highly structured machine vision applications. These applications comprise modules which can be considered as building blocks of CIM systems. Their integrated interoperation can be enabled by the emerging infrastructural tools which will be required to underpin the next generation of flexibly integrated manufacturing systems. The work described in this thesis concludes that the issues of machine vision applications and the issues of integration of these applications within manufacturing systems are entirely separate. This separation is reflected in the structure of the thesis. PART B details vision application issues while PAIIT C deals with integration. The criteria for next generation integrated machine vision systems, derived in PART A of the thesis, are extensive. In order to address these criteria and propose a complete architecture, a "thin slice" is taken through the areas of vision application, and integration at the lifecycle stages of design, implementation, runtime and maintenance. The thesis describes the reference architecture, demonstrates its use though a proof of concept implementation and evaluates the support offered by the architecture for easing the problems of software change.