| EPSRC Reference: |
GR/J65372/01 |
| Title: |
REPRESENTATION AND CONTROL IN ACTIVE VISION |
| Principal Investigator: |
Murray, Professor DW |
| Other Investigators: |
|
| Researcher Co-Investigators: |
|
| Project Partners: |
|
| Department: |
Engineering Science |
| Organisation: |
University of Oxford |
| Scheme: |
Standard Research (Pre-FEC) |
| Starts: |
01 December 1993 |
Ends: |
29 February 1996 |
Value (£): |
184,114
|
| EPSRC Research Topic Classifications: |
|
| EPSRC Industrial Sector Classifications: |
|
| Related Grants: |
|
| Panel History: |
|
|
Summary on Grant Application Form |
|
In an earlier project, we built a head/eye platform, real-time vision system and gaze controller which was capable of reacting to motion in the scene. This project was designed to give the system the ability to engage in task-oriented visual activity over extended periods, by:(i) recovering 3D motion trajectories and(ii) by recovering and describing the structure of tracked objects.Progress:The project, now early in its second year, has two parts. The first was to explore methods of recovering 3D trajectories of moving objects in the scene, using both monocular and stereo vision.To recover 3D trajectories monocularly we have developed a method of calibrating the gaze angles of the head on the assumption that the motion occurs in a plane (typically the ground plane). This work has demonstrated that an active system can act as a wide (indeed, omni-directional) field-of-view projective device, using the rotation centre of the mechanical device as the centre of projection, exploiting the one-to-one mapping between the gaze sphere and an contrived plane of projection. We have demonstrated the recovery of trajectories from both laboratory and outdoor scenes, and have explored the use of the Interacting Markov Model to describe and predict a variety of simple motion trajectories, such as straight line constant velocity, straight line constant acceleration and turn trajectories over extended periods. We have also completed installation of a stereo channel within the vision system, to recover arbitrary 3D trajectories, and have demonstrated the improvement in prediction and hence tracking performance using 3D rather than 2D trajectories. This work has developed significant new theory for affine transfer between stereo frames using multiple point features, a development of earlier work in a single camera using a minimal point set. This method of tracking via affine transfer also allows the recovery of affine structure. This has been demonstrated to be stable over a number of frames, but the structure is at present lost when there are insufficient data in even a single frame time. Current work is exploring the maintenance of the structure over periods when there are insufficient data, an essential precursor to work on recognition of the recovered structure. Other ongoing work is exploring the use of the gaze sphere and its mapping onto the contrived fronto-parallel plane to allow resumption of visual tasks after interruption. Another part of this project was to update our vision system. Rather than use T800 Transputers whose bandwidth and processing power is now limiting, we have constructed a more portable system using C40 DSPs .
|
|
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.ox.ac.uk |