الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 85 |  |  |
| | | from | 85 | | |
Abstract
Visual Surveillance is concerned with the utilization of modern digital cameras to achieve common surveillance goals. Thanks to the wide availability of low cost and highly efficient sensors, there has been a proliferation of deployments of visual security systems as seen in banks, airports and train stations. High end visual sensors also play a crucial role in autonomous robots like unmanned aerial vehicles. In this thesis, we study two variants of the visual surveillance problem.In the first problem, surveillance video streamed from a top-view camera is processed to control the orientation of multiple pan-tilt-zoom cameras in order to cover as many targets as possible at high resolutions. The problem of covering a set of static targets with a set of cam- eras is a planar variant of the classical combinatorial set cover problem and has been shown to be computationally intractable. We develop two new heuristics, compare them to existing solutions and show their superiority by extensive simulations. Moreover, to demonstrate the applicability of the proposed methods, we build and evaluate a real surveillance system for indoors pedestrian tracking. In the second problem, we study a variant of the classical pursuit-evasion game, in which an agent moving amongst obstacles is to be maintained within sight by a pursuing robot.We design an efficient algorithm that decides if the evader can take any moving strategy to
hide from the pursuer and win the game. For situations where the evader cannot win, we compute a pursuit strategy that keeps the evader within sight. Finally, if it is determined that the evader wins, we compute its optimal escape trajectory and the corresponding optimal pursuit trajectory. We analyze the algorithm, present several optimizations and show results for different environments.