Project: Monitoring of High Occupancy Vehicles
Reference: STP 14/6/13
Last update: 27/07/2004 11:20:37
This project will develop a combination of optical , electronic and electro-mechanical techniques that will determine the number of occupants in a car so that the lane restriction can be enforced effectively.
Clear imagery of a motor vehicle will be obtained such that the interior of the vehicle is clearly illuminated and views both the front and side windows acquired. Pattern-recognition techniques will be used to detect head and shoulder shaped regions to determine the number of occupants in the vehicle.
Pennard House, Station Road, Baldock, Hertfordshire, SG7 5RS
Cost to the Department: £97,932.00
Actual start date: 28 March 2001
Actual completion date: 30 June 2004
High Occupancy Vehicle Monitoring (HOVMON)
Author: C R Chatwin, R C D Young, P Birch
Publication date: 25/09/2003
Source: Professor C R Chatwin, University of Sussex
Summary of results
- In this project we have tackled a difficult real world image processing problem and shown that it is possible to solve such problems: frequently, image processing and object recognition is demonstrated on idealised or perfect images, with techniques that are seldom robust when applied to more unconstrained images.
The HOVMON project has developed new technology for continuing the number of people within cars to enforce high occupancy vehicle lanes. A new and robust method of vehicle window detection has been developed which uses colour segmentation techniques to overcomethe difficult and unique issues that occur with this problem (i.e. detecting almost transparent glass). Near infrared imaging has also been shown to be effective in detecting vehicle windows, which is a vitally important image segmentation requirement. Several state-of-the-art face detection algorithms have been evaluated and the neural network was found to be the most robust and suitable method for this application, its great strength being the ability to retain the system on false positive data and thus greatly reduce the error rate in difficult lighting conditions. The discovery of IR transmission windows at wavelengths around 1550nm and the fact that human skin is highly absorbent at this wavelength is a significant result ("A system for the identification of the Presence of Skin Tissue" - Patent application number - 0308266.6). Sussex University's work offers a number of solutions and solution combinations that permits a vehicle head count, even with poor image quality.
The next stage of this research will be performed by the remainder of the research collaborators funded for a further 6 months under the project. This consists of building a road side system and retaining the network for the specific camera systems and particular lighting conditions.
Departmental Assessment Status: Assessed by FIT Programme Advisory Group 15 July 2004. Allocated score 8/10 - Very good project, achieved targets and objectives, good quality research work, outputs with commercial potential.