Research Aims

​

Scientific Background 
This project aims to improve the performance of artificial intelligence-based video analytics in agriculture by developing and applying novel methods for Multi-Object Visual Tracking (MOT). Computer vision, video analytics, and AI can help monitor and improve animal welfare while improving farm productivity. Particularly important to industry is the need to monitor animal behaviour for early detection of the signs of disease. However, monitoring animals poses unique challenges for AI, including tracking with erratic and non-linear motion and coping with changes in appearance as animals grow. The performance of current systems is limited because they have been developed without due consideration of these challenges. Hence, this studentship will seek to develop improved MOT algorithms, laying the foundation for future methods for applying AI in agriculture and opening avenues for new commercial applications. 
 
Research Methodology 
We will use machine learning / deep learning in this project. The project will focus on three main challenges. Firstly, we will develop a method for multi-object tracking of large numbers of farm animals with similar appearances. To do this, we add constraints to an existing MOT framework to ensure geometric consistency. We will also develop a novel method of re-identification incorporating neighbourhood context to discriminate between individuals with similar but non-identical appearances. Secondly, we will extend our system to learn Animal-Specific Motion Models. Our tracker will be designed to learn to dynamically switch between the different learned motion models. Finally, we will tackle the problem of tracking as animals grow and their behaviours and appearances change.  
 
Training 

The successful candidate will be offered the chance to attend modules of the MSc course in Animal Behaviour and Welfare (QUB) to obtain an understanding of the challenges to be addressed. The successful student will also have the chance to attend modules on MSc in Data Analytics (QUB), especially in Machine Learning, Deep Learning and Generative adversarial networks. In both cases the successful candidate will acquire a rare combination of technical, scientific, and hands-on skills that are highly valued in both industry and academia. 

​