This project will involve all of this and provide an opportunity to participate at the forefront of exciting research working with staff in the School of Electronic and Electrical Engineering and in the School of Civil Engineering.
|
|
|||||
|
Are
you interested in developing novel applications of computers, working closely
with medical staff on medical applications of computers, computer vision
research. Do you thrive on a challenge ?
This project will involve all of this and provide an opportunity to participate at the forefront of exciting research working with staff in the School of Electronic and Electrical Engineering and in the School of Civil Engineering. |
||||
| David Pycock | Tel.: | +(0)121 414 4330 | |||
| Email: | D.Pycock@bham.ac.uk | ||||
| WWW: | http://www.eee.bham.ac.uk/pycockd | ||||
| Background | |||||
| Frequent inspections are needed to check the presence and visibility of traffic signs and highway furniture. Manual surveying normally requires several passes through a highway for all the necessary checks for a comprehensive survey to be completed. This is costly and logistically difficult to organise. All the systems described are concerned with traffic sign interpretation as a driver aid and not with inspection. Inspection requires not only that the sign is identified but also that tits integrity is assessed; this is a more demanding task and requires a greater degree of robustness. | |||||
| Objectives | |||||
 |
To investigate colour-based criteria for robust motion tracking. | |
|
To develop robust colour-based object models for traffic sign recognition and inspection that will work in highly cluttered scenes. | |
|
To identify occluded traffic signs. |
| Previous Research |
| Previous research into the detection and recognition of traffic signs has concentrated on the use of fixed monochrome cameras. Piccioli et al describe work using colour images to rapidly identify regions of interest. They describe a complex, integrated scheme for identifying signs which achieves high correct detection rates (93%), relatively low false detection rate (14%) with difficult images and takes 15 sec, on average, to process edges from a region of interest. The method described combines data from a series of images but uses little knowledge about the appearance of road signs to guide interpretation. |
| Previous research on model-based schemes of object recognition has used monochrome images and sought to accommodate biological variation and variation in the appearance of 3-D objects in 2-D projections. In the recognition of road signs neither of these patterns of variation are important. Rather we need to accommodate occlusion. Previous research in the School of Electronic and Electrical Engineering has shown how 2-D and 3-D objects can be recognised from line drawings with high degrees of occlusion. We have also developed robust schemes for monochrome image interpretation and we propose to adapt both techniques to use colour images. |
| Proposal | ||
| We propose
to use colour images and appearance models at all stages of the traffic
sign interpretation process to improve robustness without incurring speed
penalties and to use an articulated camera system so that regions of interest
can be tracked and imaged in more detail.
The recognition of damage to signs will require an ability to: |
||
|
Identify posts and the absence of a sign. | |
|
Identify sign supports on a post. | |
|
Particular appearances such as an illuminated sign with no face panel. | |
|
Recognise the presence of graffity. |
| Tracking will initially be evaluated from wide angle recordings of natural scenes. Once a good a level of performance has been achieved with this we will evaluate a closed loop system from video recordings. |
| Facilities and Benefits |
| Extensive computing resources, a well stocked library and a colour active vision head (see top of page) are available to support this research which is expected to lead to publications in international journals. There will be an opportunity to submit results for presentation at national and international conferences. The analytical and computing skills that you will develop are much sought after by potential employers. |
| Contact
Information
Mrs J. Squire
|
|
|||
+(0)121 414 4292