Welcome to the web site of James Alexander Grove Annesley
Use the grey "menu" patches labelled 'PhD' and 'Publicity' above to navigate.
I work at Kingston University, London. I am currently finishing my PhD here which is about visual surveillance meta-data. This web site provides an introduction to the PhD and some other publicity material.
The information in the PhD section is taken from an award winning writing piece, written by me, that jointly won the 2005 Imaging and Sensing Faraday writing competition. For more information on my work, please go to the links and choose the Digital Imaging Research Centre (DIRC) link.
Thank you
PhD Introduction
My PhD is called GENERICK - (Generation, Encoding and Retrieval of CCTV-derived Knowledge)
Closed-circuit television (CCTV) surveillance is big in Britain - the average person is filmed many times each day going about their daily lives. What is the purpose of it? Is this intrusion into our private lives justified?
Current CCTV state of Art
These days, the police often waste hours trawling through recorded footage of video tapes. They are looking for evidence in an effort to investigate and, indirectly, to deter crime. In the Brixton nail bomb attack investigation for example, hours and hours of CCTV footage were searched. In this particular case, Brixton's large number of well installed CCTV cameras were invaluable in convicting the perpetrator, but all too often the reality is different due to badly sited cameras or grainy image quality caused by archaic video recorders, worn tapes or a host of other reasons. Some of these problems can be addressed by replacing existing analogue CCTV systems with digital technology.
Future CCTV
Digital CCTV surveillance systems introduce the possibility of improved image quality, incorporating computer processing, easier installation, better data transmission and, in some cases, a level of artificial intelligence, e.g. using triggers to record 'events', such as cars driving into car parks, or more exotically, the violation of electronic 'tripwires' (just think of Tom Cruise in Mission Impossible). As well as causing immediate alarms, these video events can be stored in a database to allow a quick review and additional information supplied by machine vision algorithms can be stored too. We call this information 'metadata' and it can provide abstract information about a scene, such as the presence of objects and any characteristics of these objects. For example, a robbery is committed and a witness gives a description, the colour of the criminal's clothes can be used to narrow down likely suspects.
Metadata
Video metadata can be described by the using colours in the video, so long as they produce descriptors of high variance when compared together. Work has been done assessing the validity of a average colour representation. This representation involves calculating the average colour over all video frames in a video sequence. The standard deviation is calculated over the frame means. This is done for each sequence in a test data set and the results showed little variation between sequences. Further work is being done to produce descriptors based upon the dominant colours in the video. The dominant colours are produced through a histogram of the video and a clustering algorithm to separate the strongest n colours.
Metadata in Surveillance
Adding a bit of intelligence to the data gathering process throws up all sorts of interesting options, including the ability to be a bit more selective about what data to record, for example not storing endless images of nothing happening. It is far better - from a storage point of view - of only recording moving events. Movement is detected by signal processing and, using a model of a background scene, an intelligent CCTV system can understand when significant foreground movement occurs. Metadata is generated by removing the background and analysing the extracted information. In other words, a description of information that people automatically understand, such as an image's colour makeup, is extracted to form a grammar that a computer can interpret. A mathematical procedure generates the values and these are stored as text, and this metadata - along with time stamp and event identification - is stored in a database. Example of use this metadata include quick browsing, event-based search ('who walked through this doorway in the last four hours') and similarity matching ('have other events like this happened recently?').
MPEG in Surveillance
A standard way of defining such metadata addresses problems of compatibility. The metadata standard MPEG 7, designed by the Moving Pictures Expert Group, when used in conjunction with the MPEG 21 standard can mean the secure delivery of compatible metadata descriptions over computer networks, such as the Internet. Both contain descriptors for protecting the integrity of the digital data, which is otherwise easy to copy and adulterate. The defined meta-data might be based upon the MPEG7 standard.
Other Surveillance Research
Other methods used in surveillance are object tracking, biometrics to automatically identify people, cooperative camera networks and neural networks. This research is compatible with all such approaches and will contribute to the development in this field.
Implications of the Research
Implications of successful implementation of this technology are potentially huge. Consider alarms delivering, possibly via mobile Internet or telephone networks, information on the location and the colour signature of objects. Police could review stored data ore quickly, not only browsing by events but also using stored metadata, i.e. looking for particular characteristic, such a suspect wearing a red shirt or black trousers.
Conclusion
Contrary to what one might believe from having seen Hollywood's Enemy of the State, there is much to achieve before such systems are in everyday use: problems of low-light levels, colour constancy between cameras, occlusions of foreground objects and changing conditions in an outside environment must still be addressed. One must also consider the many complex - but not insurmountable - implementation issues that would go hand-in-hand with bringing this technology to the market, such as data protection, security and transmission. Then there is also the question of whether the public will accept the use of this technology... anyone for a date with big brother?
MTBF Timer
The MTBF Timer is a program written by yours truely that clocks the computer's up-time. It is hosted on the International CNET web site for shareware and freeware software.
Look Away
Worried about eye-strain when looking at the computer screen? Try `Look Away` - a little program I wrote. It warns you every few minutes to look away from the screen. It comes complete with an installer/ uninstaller.
Journal Papers
1. R Aveyard, BP Binks, JH Clint, PDI Fletcher, T.S. Horozov, B Neumann, VN Paunov, J. Annesley, S.W. Botchway, D Nees, AW Parker, ADW Ward, "Measurement of Long-Range Repulsive Forces between Charged Particles" in 'Physical Review Letters', 88(24) pp. 246102. (2002)
2. R Aveyard, BP Binks, JH Clint, PDI Fletcher, B Neumann, VN Paunov, J. Annesley, S.W. Botchway, AW Parker, ADW Ward, AN Burgess, "Drag Forces on a Stationary Particle in Flowing" in 'Langmuir', 18(24) pp. 9587 -9593. (2002)
Conferences
3. J. Annesley, J. Orwell, "On the Use of MPEG-7 for Visual Surveillance", 6th IEEE International Workshop on Visual Surveillance, May, Graz, Austria, (2006)
4. J. Annesley, V Leung, S.A. Velastin, A. Colombo, J. Orwell, "Fusion of Multiple Features for Identity Estimation", International Conference on Imaging for Crime Detection and Prevention (ICDP), Visual Information Engineering IET, June, pp. 534-439. ISBN/ISSN 086341647/0537-9989(2006)
5. J. Annesley, J. Orwell, J.R. Renno, "Evaluation of MPEG7 Color Descriptors for Visual Surveillance Retrieval", Joint IEEE International Workshop on Visual Surveillance and Performance Evaluation, VS-PETS Beijing, China, (2005)
Technical Reports
6. J. Annesley, "MPEG-7: Two Useful Restrictions for Visual Surveillance" in DIRC-tr-2007-02 1, (2007)
|