EPSRC logo

Details of Grant 

EPSRC Reference: GR/H77668/01
Principal Investigator: Murray, Professor DW
Other Investigators:
Zisserman, Professor A
Researcher Co-Investigators:
Project Partners:
Department: Engineering Science
Organisation: University of Oxford
Scheme: Standard Research (Pre-FEC)
Starts: 01 October 1992 Ends: 31 March 1996 Value (£): 254,176
EPSRC Research Topic Classifications:
Image & Vision Computing
EPSRC Industrial Sector Classifications:
Related Grants:
Panel History:  
Summary on Grant Application Form
This project is investigating the use of projective invariance in computer vision to enable navigation without the explicit recovery of metric Euclidean 3D data. The project, now in its final year, has three phases:(i) the development of visual navigation and path planning methods in non-metric space(ii) the development of a lightweight active stereo system and(iii) the application of (i) and (ii) to navigation on a mobile vehicle. Progress:Two robust working vision systems have been implemented. The first is a single camera mounted on a robot arm which is able to navigate between obstacles in an unknown scene to reach a specified target position. The second is an active stereo head mounted on a mobile vehicle which is also able to manoeuvre between obstacles in an unknown scene. The original contributions of the work are enumerated in detail below.1. Original methods for setting up a 3D affine coordinate frame have been developed, for unknown camera calibration and constrained camera motions. The first method is partly based on existing vision work and works with a camera undergoing pure translation, while later work is completely novel and more general, working with a camera undergoing translation in a plane with rotation around an axis perpendicular to the plane. The latter is a very natural exploratory motion, whether for cameras on a mobile vehicle or humans. We have also used approximately or partly known camera calibration, whenever available, in order to determine a more well-conditioned choice of affine coordinate frame. 2. A scheme has been implemented to initialise newly appeared 3D structure and updating existing structure over time, working in an coordinate frame. 3. A novel method has been devised for the computation of areas of unoccupied space around the camera, coupled with methods for selecting the most appropriate route for the robot or AGV through the space, using affine constructions in an affine coordinate frame. 4. The above methods have been integrated to provide a global view of the 3D structure and areas of unoccupied space in the surrounding environment, with a vision process providing close attention to the most proximal (and therefore dangerous) obstacles when passing them. This work addresses problems of integrating and passing information between different visual functions in a single system. 5. A robust matcher has been developed for image points, one capable of dealing with noisy and incorrect data. Matching is a multi-stage process, removing incorrect matches and adding new ones at each stage according to the constraints generated from the data itself. Mismatched image features are a difficult problem for any system which functions over an extended time period, since they can lead to accumulating degradation in performance. We transform image data appropriately to obtain the best possible conditioning for solution of linear systems, and have employed techniques from robust statistics to identify and remove mismatched features. The most recent work is extending the matcher to handle lines as well as points. 6. All of the above have been incorporated in a stereo system which makes full use of the constraints available. Stereo improves functioning in various ways, by offering a wide baseline for computation of new structure, and providing more powerful checks in the feature matching process.
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
Description This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
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