- An Image-based Facial Animation
This project is to develop an image-based animation system for facial
expressions by combining computer vision and computer graphics
techniques. A real-life actor's facial expressions will be captured
on images, and the facial mash model of the actor will be generated by
Poser (a 3D character modelling and animationprogram by Curious Labs).
Objectives of this project include: analysing motion within the
actor's face using a computer vision technique; mapping the image
locations of the detected motion to the facial mesh; and animating
facial expressions on computer graphics by interpolating and rendering
the detected motion from the images.
To detect motion, a computer vision technique such as the optical flow
will be employed to calculate motion vectors. Graphical animation of
the facial expression requires understanding of the Facial Action
Coding (FAC) parameters that defines muscle regions for facial
expressions in order to accurately map the detected motion to the 3D
face mesh model. Animation may be achieved by an existing
interpolation module of the Auslan Tuition System that was previsouly developed
within our school.
A student who is familiar with both computer graphics and computer
vision will be suitable for this project.
This is a subproject of the animation project
- Detecting Unusual Events In Surveillance Scenes
An automatic surveillance system requires detection of events that are
unusual to the known environment. Unusual events may include a sudden
change of motion speed within the scene, or detection of suspicious
motion of an individual.
This project is to develop a vision-based automatic event recognition
system for surveillance applications. Objects in the scenes need to be tracked and analysed to extract temporal changes of their appearances such as the locations and shapes. This information is then used to recognise suspicous events.
Tracking and shape detection algorithms are already implemented by an honour student in 2004 for an automatic human detection project. This project may adapt these techniques to reliably extract image features for this particular application. An important part of this project is to devise a technique to classify those image features to determine unusual events. This process involves recognising multi-agent events using temporal logic.
A student with a good mathematics background is suitable for this project, regardless of the familiarity with computer vision. This project will be co-supervised with Ass. Prof. Mark Reynolds.
- Human Pose Detection
This project is to develop a technique to detect specific human poses from a live-video. Such task is important for automatic surveillance systems where suspicious (target) body poses, for example, gun-pointing or hold up poses, can provide a cue to alarm the security personnel.
The system will use live-input images captured from a camera attached to a PC. The images will then be analysed by extracting features to represent body poses appearing in each image frame, and classifying the extracted features to detect target poses.
Windows programming experiences will be useful.
Return to the list of 4th year projects