Contributed by: Rafal Zbikowski, r.w.zbikowski@cranfield.ac.uk

'Bioinspired omnidirectional vision: Integrated approach' PhD project to commence in June 2012

Hosting Institution: Centre for Autonomous Systems, Cranfield University, England

www.cranfield.ac.uk/cds/autonomous/index.html

Supervisor: Professor Rafal Zbikowski

r.w.zbikowski@cranfield.ac.uk

Project summary

The aim of this three-and-a-half year PhD project is rigorously to analyse the capabilities of bioinspired omnidirectional vision based on a spherical sensor, mounted on a moving platform, and validate the analysis with a software simulation tool. A key aspect of the insect visual perception of motion is simultaneous use of omnidirectional perception of geometric cues (lines and simple figures with their orientations and symmetries) and spherical optic flow (global relative motion, object boundaries through vector field discontinuities). The spherical eyes of flying insects allow a global view of both the static and dynamic aspects of the environment together with a comprehensive perception of the insect's motion relative to the environment.

Geometric and optic flow cues are mapped onto a sphere so that each object and its relative motion are localised by appearing on the relevant part of the sphere.

The spherical perception of insects is based on matching simple patterns and their orientation by comparing the perceived objects and their relative motion with a small number of geometric and optic flow primitives. The static and dynamic primitives are matched locally on small patches of the sphere and then matched globally on overlapping hemispheres so that local and global perception of objects and relative motion is achieved simultaneously. The integration of the local and global overlapping imagery enables robust perception and aids disambiguation of patterns, orientation and motion. The above important biological insights are qualitative in nature and their use for bioinspired engineering requires rigorous quantitative analysis.

The proposed analytical work will be supported and validated by a software tool whose development will be an essential part of the project. The Matlab-based tool will allow simulation of a spherical sensor moving in an indoor environment (confined by walls).

That environment will be populated by static or moving obstacles which can be solid or have holes, the latter in order to investigate the ability to recognise openings for potential pathways through the obstacles. The spherical sensor will traverse the environment using trajectory planning and the resulting sensor output (geometric cues and optic flow) will be available for analysis and validation.

This three-and-a-half year PhD project is sponsored by the UK Ministry of Defence (MoD) under the Dstl National UK PhD programme and is open to EU nationals only. The MoD funding includes a student stipend of £13,590 per annum and is available from June 2012. For further information or to apply, please contact the project 

Supervisor, Prof Rafal Zbikowski, r.w.zbikowski@cranfield.ac.uk by 30 June 2012.