Projects Profile
Project Title
VISTA: Video Analytics Software for Real-Time Human Profiling
Partnership
Prof. Shaogang Gong, Queen Mary, University of London
Prof. Yunhong Wang, Beihang University
Project Aim
The aim of the project presented in this ICUK application is to develop a software system capable of automatically searching and profiling the gender, age and type of human subjects/populations captured in video, and to provide a facility for either real-time information gathering on people movement, or off-line demographic analysis on human populations in public or private space.
Inspiration for the projects
There are growing numbers of CCTV cameras for surveillance purposes. The bottle neck for such systems is on the storage and analysis of the video footage that is collected. The existing approach is manual and retrospective - after the event, a human operator retrieves the relevant video clips/frames by hand, which is expensive, prone to human error and doesn’t capture the event at source. Moreover, due to limited storage capacity, video data have to be deleted after a short period of time.
To solve the above problem, QMUL and Beihang University have developed a technology that enables the automatic identification of events as they occur in real time.
Commercial Potential and Further Development
The software has various applications in security, military and retail management. The support of the ICUK PoC fund will enable the researchers to build up a demonstrate product to attract customers and licensees.
Project Activities
The project is supported by £70,000 from the ICUK development fund and RMB 450,000 from the Chinese National High Technology Research and Development (836) Program. A full time RA in the UK and 5 research students in China will work under the guidance of the PIs to develop a demonstrator prototype. Specifically, the following activities will be carried out:
1. To design an integrated software architecture
2. To implement a code-base in C++
3. To optimize algorithm parameters for variable lighting conditions and crowding
4. To design and implement a user interface and visualization component
5. To test and evaluate the system against realistic data from public scenes
