This thesis presents a method for mobile robot localisation. Both the theoretical and practical aspects of the problem in general and our specific solution are considered. Chapter 2 presents a general discussion of existing solutions and other related work. The model of visual attention that is employed for feature extraction is presented in Chapter 3. Chapter 4 presents our method for learning (tracking) landmarks for the purposes of localisation and Chapter 5 presents the online method for employing the learned landmarks in order to obtain a position estimate. Finally, the experimental results presented in Chapter 6 demonstrate the robustness of the method under a variety of environmental conditions. Chapter 7 concludes with a discussion of the experimental results and possible directions for future work.