| EPSRC Reference: |
GR/A00010/01 |
| Title: |
AF: EXTENTIONS OF RELATIONAL GRAPH MATCHING |
| Principal Investigator: |
Wilson, Professor R |
| Other Investigators: |
|
| Researcher Co-Investigators: |
|
| Project Partners: |
|
| Department: |
Computer Science |
| Organisation: |
University of York |
| Scheme: |
Advanced Fellowship (Pre-FEC) |
| Starts: |
01 April 2000 |
Ends: |
30 September 2003 |
Value (£): |
96,363
|
| EPSRC Research Topic Classifications: |
|
| EPSRC Industrial Sector Classifications: |
|
| Related Grants: |
|
| Panel History: |
|
|
Summary on Grant Application Form |
|
This proposal details a programme of work aimed at developing a statistical framework for modelling relational structures and the impact of these structures on other levels of abstraction. The programme builds on the author's published work which has established the foundations for a simple model for matching and editing relational structures. The aims of the proposed research are to enhance the capabilities of the existing work by developing more sophisticated statistical models and by incorporating facilities for learning. The improved statistical modelling will focus on the issue of how to couple relational models to other processing levels in representational and perceptual hierarchies. By incorporating the facility for learning the author aims to enhance the adaptive capacities of relational models. In particular, the aim is to overcome some of the difficulties traditionally associated with the use of relational models.There are in fact a number of difficulties that limit the practical utility of relational representations. These limitations revolve around the difficulty of extracting reliable graph primitives, generating salient and robust relations between these primitives and controlling the inevitable noise and clutter in the resultant graphs. Recent work by the author has gone some way towards addressing this last issue. By providing a finely graded measure of relational consistency, graphs can be matched both with poor initial conditions and with significant quantities of corruption. In addition, the divergence between graphs before and after edit operations can be used to identify and remove spurious elements from graphs under match. The intention is to develop elements of this work to address the limitations associated with the use of relational graphs. The identification of graph nodes (representing objects in a data-set) can be enhanced by appealing to high-level models of possible structures in the data. If these models are cast in terms of relational graphs, the requirement becomes one of coupling relational abstractions to the low level processing of raw data. This observation leads to the first element of this proposal; there is a general requirement to couple relational graph processes to other data interpretation steps and other levels of processing in order to increase the utility of these methods. One aspect of this proposal is therefore to develop a general statistical framework that allows the coupling of the relational level of abstraction to other processes.The second limitation in graphical approaches consists of the problem of identifying relational structures which are both salient and robust. The approach in this case will be to adopt a learning framework based around the graph-edit methods developed by the author. In this framework, information theoretic measures will be used to quantify both the quality of a relational structure and the cost associated with different edit operations. 'These measures, when combined with an appropriate learning methodology will allow useful relational structure to be learnt.Finally, these methods will be assessed in the context of vision algorithms by implementing methods for segmentation, perceptual grouping and object recognition.
|
|
Key Findings |
|
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
|
|
Potential use in non-academic contexts |
|
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
|
|
Impacts |
|
| Description |
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk |
| Summary |
|
| Date Materialised |
|
|
|
|
Sectors submitted by the Researcher |
|
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
|
| Project URL: |
|
| Further Information: |
|
| Organisation Website: |
http://www.york.ac.uk |