Overview
Topic status: We're looking for students to study this topic.
Project Summary
A common problem in security and law enforcement is locating a person in a crowd or a large space. This is typically done by a manual search of the environment, either on foot, using CCTV footage or a combination of both. This approach is very labour intensive and not ideal when appropriate when time is critical.
A baseline system has been developed for locating a person from a semantic description, however it is limited in that only height, torso and leg colour are used as traits, and these traits are weighted equally when conducting a search. Additional traits are required to improve the search performance, as well as an intelligent fusion process that is able to combine traits based on the confidence of the detection result, and the reliability of the trait itself.
Expected outcomes, applications and/or benefits
- An understanding of image segmentation, object detection and feature fusion methods.
- A multi-camera database suitable for evaluating semantic person search techniques.
- An improved person localisation system where a semantic query is used to initialise the search.
Required student skills/experience
Strong C++ programming skills and experience in image processing.
- Study level
- Vacation research experience scholarship
- Supervisors
- QUT
- Organisational unit
Science and Engineering Faculty
- Research area
- Keywords
- semantic descriptions, crowd, image, detection
- Contact
- Contact the supervisor for more information
Professor Sridha Sridharan
Associate Professor Clinton Fookes
Dr Simon Denman
