Visual Surveillance: Dynamic Behavior Analysis at Multiple Levels

New cameras are installed daily all around the world. Hence it becomes increasingly important to develop methods for reducing the manual effort that is still required for video analysis. This book presents several methods to automatically analyze the behavior of dynamically moving persons in real-world scenes. Because an image captures different aspects of a scene depending on the scale level, the analysis at different scale levels requires individual approaches to explore the available kinds of information. The innovations include a method to find faces and estimate their pose, a multi-person tracker for complex scenes, an algorithm to analyze the activity of interacting agents, a technique to detect novelty from webcam footage, and a method to estimate the structure of a scene.

• Type: Dissertation
• Year: 2009
• Authors: Michael D. Breitenstein
• Keywords: Computer Vision, Detection, Machine Learning, Scene Analysis, Tracking, Visual Search