Many machine vision systems today are more than pure inspection systems, as they make it possible to recognise trends in production processes early on often using intelligent reasoning of the complete production process. Intelligent reasoning is achieved via computational intelligence which deals with the development of algorithms inspired by human cognition. In simple terms, computational intelligence seeks to develop models that can reason, understand or learn like a human. The ability to spot patterns, adapt to new and unusual data, and to be robust to non-perfect data are hallmarks of computational intelligence methods and are all issues encountered in current industrial processes.
Computational intelligence alongside machine vision can deliver both the primary inspection information in early stage manufacturing and also comprehensive and important secondary information relating to the production process. This combination of machine vision and computational intelligence dovetails very nicely with the requirements of an automated inspection system - to spot trends in past data and identify changes in the future; automated inspection systems require a model which can adapt to changing product designs in a controlled, easy to understand fashion; and a model which will not completely fail due to potentially noisy data/outliers.
Machine vision can include many methods however the dominant approach found within industry but often a template matching approach where a pixel-to-pixel comparison of two images is performed is applied. This technique is especially useful when the object's surface or object's shape is very complex and when finding defects like smudges, scratches etc. The reference image is a previously prepared image which is used to compare with images from the camera. This technique allows quick comparison inspection but some specific conditions must be met such as stable light conditions, position of the camera and the object must be static, precise object positioning, and camera calibration. Additionally a significant problem with this technique is that even small changes in product design require the preparation of an updated reference image which takes a significant amount of time given frequent product revisions.
This project will develop an improved suite of machine vision technologies suitable for the automated inspection of wafers. The reference image approach will be replaced by alternative image processing techniques based on local features combined with computational intelligence techniques so that changes in the product design will not require an updated template for the detection of defects. Rather than performing pixel by pixel comparisons between complete images, the system will be able to detect localised image features that correspond to a particular defect type. In addition, the intelligent reasoning system will be capable of monitoring existing and detecting new defects and image changes that occur as a result of process design changes, equipment wear or which are a result of earlier processes. This will enable the system to maintain continuous service even new or update products are developed.
If the University receives a large number of applicants for the project, the following desirable criteria may be applied to shortlist applicants for interview.
This project is funded by:
The scholarship will cover tuition fees at the home rate and, for applicants with UK residence only, a maintenance allowance of not less than £15,480 per annum for three years. EU residents may also apply but if successful will receive fees only.
As Senior Engineering Manager of Analytics at Seagate Technology I utilise the learning from my PhD ever day
Adrian Johnston - PhD in InformaticsWatch Video
Monday 19 February 2018
When applying for this PhD opportunity please quote reference number: