The aims of this course are to introduce the principles, relevant methodologies and applications of automated visual surveillance systems, as well as to give an overview on the state-of-the-art and open challenges. The course will cover relevant issues from both methodological and application points of view: (1) Algorithmic concepts: motivating issues, building blocks of surveillance systems such as object detection, spatial grouping and clustering, object classification, object tracking, behaviour representation and understanding, incorporating available knowledge, performance evaluation methodologies. (2) System-specific concepts and emerging trends: distributed and embedded visual computing, the use of multiple cameras and sensors. (3) Application examples: state-of-the-art, unsolved problems and future trends. The course will also discuss applied examples from the task-oriented view clarifying what are realistic performances, providing performance measures and examples for modern surveillance systems, discussing the limits of current systems and giving a preview which improvements can be expected in further years to come. The course will be illustrated by research papers and numerous examples of our own research.

Introduction, Motivating trends, Tasks and related challenges, Methodologies, Performance evaluation, Surveillance system concepts, Application examples, Case studies and Emerging trends. Objectives At the end of the course attendants should obtain an overview on the motivating trends behind methodologies, their combinations and their role in a vision system, understand the underlying concepts of the visual processing chain from bottom-up (data-oriented) and top-down (task-oriented) perspectives within a surveillance system, learn the basic mechanisms behind its low-, mid- and high-level visual processing methods, be able to analyse the robustness, generalisability, and performance of different approaches, be able to judge the boundary between realistic and unrealistic performance, understand in depth some practical application-oriented tasks, such as pedestrian counting, tracking, and activity interpretation.

No lecture notes are available.