| EPSRC Reference: |
GR/J07884/01 |
| Title: |
A RECONFIRGURABLE PARALLEL VISUAL ARCHITECTURE THROUGH PROTOTYPE INSTRUMENTATION AND TRANSFORMATION |
| Principal Investigator: |
Wallace, Emeritus Professor A |
| Other Investigators: |
|
| Researcher Co-Investigators: |
|
| Project Partners: |
|
| Department: |
Electrical and Electronic Engineering |
| Organisation: |
Heriot-Watt University |
| Scheme: |
Standard Research (Pre-FEC) |
| Starts: |
15 November 1993 |
Ends: |
14 December 1996 |
Value (£): |
200,842
|
| EPSRC Research Topic Classifications: |
|
| EPSRC Industrial Sector Classifications: |
|
| Related Grants: |
|
| Panel History: |
|
|
Summary on Grant Application Form |
|
1. Specification and implementation of a visual system for scene interpretation using a mixture of sequential, cascaded and parallel processing of depth and intensity data.2. Development of our approach to parallel program development using functional prototype transformation and instrumentation. In particular, we propose to implement functional abstract data types as distributed processes. Progress:l. We have completed the specification of the abstract data types for the parallel processes of the depth and intensity routes at all levels. This includes both the basic type definitions for iconic and higher level data, and the associated operations or algorithms.2. We have used functional prototypes (SML) of the parallel algorithms to identify and evaluate possible options previous to final implementation in occam on the target machine (a Meiko Computing Surface). In addition, using techniques of program transformation and the correspondences between higher order functions in the prototype and parallel skeletons, we have been able to obtain predictions of the performance of visual systems using several parallel algorithms. These have been used to provide static load balancing of the final implementation. 3. We have implemented parallel code for the majority of the processes defined in the specification, including a complete parallel implementation of a pipeline of parallel intensity processes, using edge detection, contour extraction and matching by perspective inversion. Currently we are completing parallel code for surface segmentation and cooperative matching by a clustering process in pose space, using primitives derived from depth and intensity data. 4. We are extending our functional approach to include further higher order functions and appropriate data structures for a wide range of visual tasks.Although aimed a parallel vision in particular, our work on parallel program development and instrumentation has more general applicability. Within the course of the project, we wish to tackle more complex problems associated with parallel visual systems, and examine the applicability of the functional prototyping to problems of control where time, and hence order of evaluation, is a determining factor.Using depth and intensity data, we have several routes to achieve, for example, object recognition or location. Hence, we can examine more fully issues of dynamic control of the parallel system. This should include control of the module parameters to obtain optimum accuracy of object location, for example, but also control of the resources of the parallel machine to achieve the optimum efficiency. Allocation of system , as opposed to module resources should be opportunistic, rather than pre-determined by static analysis.
|
|
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.hw.ac.uk |