Copy the page URI to the clipboard
Gill, Kapul Dev
(1991).
DOI: https://doi.org/10.21954/ou.ro.0000fce9
Abstract
With the growth of industrial automation, it has become increasingly important to validate the quality of every manufactured part during production. Until now, human visual inspection aided with hard tooling or machines have been the primary means to this end, but the speed of today's production lines, the complexity of production equipment and the highest standards of quality to which parts must adhere frequently, make the traditional methods of industrial inspection and control impractical, if not impossible.
Subsequently, new solutions have been developed for the monitoring and control of industrial processes, in realtime. One such technology is the area of machine vision. After many years of research and development, computerised vision systems are now leaving the laboratory and are being used successfully in the factory environment. They are both robust and competitively priced as a sensing technique which has now opened up a whole new sector for automation.
Machine vision systems are becoming an important integral part of the automotive manufacturing process, with applications ranging from inspection, classification, robot guidance, assembly verification through to process monitoring and control. Although the number of systems in current use is still relatively small, there can be no doubt, given the issues at stake, that the automotive industry will once again lead the way with the implementation of machine vision just as it has done robotic technology.
The thesis considered the issue of machine vision and in particular, its deployment within the automotive industry. The thesis has presented work on machine vision for the prospective end-user and not the designer of such systems. It will provide sufficient background about the subject, to separate machine vision promises from reality and permit intelligent decisions regarding machine vision applications to be made.
The initial part of the dissertation focussed on the strategic issues affecting the selection of machine vision at the planning stage, such as a listing of the factors to justify investment, the capability of the technology and type of problems that are associated with this relatively new but complex science.
Though it is widely accepted that no two industrial machine vision systems are identical, knowledge of the basic fundamentals which underpin the structure of the technology in its application is presented.
This work covered a structured description detailing typical hardware components such as camera technology, lighting systems, etc... which form an integral part of an industrial system and discussions regarding the criteria for selection are presented. To complement this work, a further section is specifically devoted to the bewildering array of vision software analysis techniques which are currently available today. A detailed description of the various techniques that are applied to images in order to make use of and understand the data contained within them are discussed and explored.
Applications for machine vision fall into two main categories namely robotic guidance and inspection. Obviously within each category there are many further subgroups. Within this context the latter part of the thesis reviews with a well structured description of several industrial case studies derived from the automotive industry, which illustrate that machine vision is capable of providing real time solutions to manufacturing based problems.
In conclusion, despite the limited availability of industrially based machine vision systems, the success of implementation is not always guaranteed, as the technology imposes both technical limitations and introduce new human engineering considerations.
By understanding the application and the implications of the technical requirements on both the "staging" and the "image-processing" power required of the machine vision system. The thesis has shown that the most significant elements of a successful application are indeed the lighting, optics, component design, etc... - the "Staging". From the case studies investigated, optimised "staging" has resulted in the need for less computing power in the machine vision system. Inevitably, greater computing power not only requires more time but is generally more expensive.
The experience gained from the this project, has demonstrated that machine vision technology is a realistic alternative means of capturing data in real-time. Since the current limitations of the technology are well suited to the delivery process of the quality function within the manufacturing process.