With appropriate adjustment, any of the projects below could be suitable for a BE(SE) final year project (12 points), an Honours Research Project (24 points), or a MSc project (24 points). If you have in mind a computer vision research topic that is not listed below, I would be interested to hear from you. You are also strongly advised to take the CITS4240 Computer Vision unit offered in the first semester if you haven't already done so.

1. Reconstructing the Path, Body and Possibly Head Orientation of Navigating Ants
   

   Bull-ants navigate during the day or night over tens of meters to and from the nest. We are interested in what they see along the way and what visual cues they use to make navigational decisions. To do this, we have to accurately reconstruct their path, body orientation and if possible their head orientation. Some Matlab code has been written already that automatically tracks the ants and we can work out body and head orientation from that information. To do this, we glue retro-reflective tape onto the animal's abdomen and head and then illuminate the scene with infrared. Infrared cameras then pick up the signal, which at night in particular gives us great contrast.

   The aim of this work is to be able to follow ants along their path with a video camera and then later stitch all the images (or the coordinates of the ant back together to reconstruct the full path. We can currently do this only by having a grid marked on the ground. It should be possible to do away without the grid and use image information alone to reconstruct the path. The outcome would be an extremely flexible tool to reconstruct what animals do over extended distances. To be useful, the system does not have to operate without ground-markers altogether, but the fewer markers we need, the better it would be. The camera might need to be stabilised if hand held, or could be mounted on a tripod and then the tripod moved as necessary.

   This project will be jointly supervised with Associate Professor Jan Hemmi (School of Animal Biology) and Associate Professor Ajmal Mian (School of Computer Science and Software Engineering).

---

2. Automatic Tracking of Fiddler Crabs

   Much of the work carried out by Associate Professor Jan Hemmi, a co-supervisor of this project, is based on tracking the movements of fiddler crabs as they move around the field of view of a stationary video camera. The crabs are very small and quite difficult to see (they are maybe 8-10 pixels long) and have articulated claws and legs. Some Matlab code has been written that uses image matching to track the crabs, but it does not work un-supervised. A number of different ways of doing this have been attempted, but image matching so far is clearly the best.

   As we do not have a good model crab, and rely on a previous image of the video, there is drift in the tracking (every time we go to the latest image template, we consolidate our tracking error). The challenge would be to find a framework that allows more robust tracking. It does not have to be fully automatic, so long as the software can flag where the results are unreliable, so we can manually check. Also, crabs can disappear down into their individual burrows, so if the tracking could detect when crabs go down and when they come back, that would be a big step. Movie sequences are often 1-2 hours in length.

   This project will be jointly supervised with Associate Professor Jan Hemmi (School of Animal Biology) and Associate Professor Ajmal Mian (School of Computer Science and Software Engineering).

Return to the 4th year project list