M.Sc. Tezi Görüntüleme

      Automated video surveillance has emerged as an important research topic in the computer vision community. The growth in this area is being driven by the increased availability of inexpensive computing power and image sensors, as well as the inefficiency of manuel surveillance and monitoring systems. Applications such as event detection, human action recognition, and semantic indexing of video are being developed in order to partially or fully automate the task of surveillance. To be succesful, such applications require real time motion detection and tracking algoritmhs, which provide low-level functionality upon which higher level recognition capabilities can be built.

      This work presents a new fast background modeling and maintanence technique for real time visual surveillance system for tracking and implementing people activities in indoor and outdoor environments. Motion objects are detected based on background model initialization with statically approach, and classified using a skeleton generation. Human is tracked using trajectory maps obtained from video sequences. The motion is also analysed with periodic variations in this skeleton structure. Experimental results demonstrate the robustness and real-time applicability performance of the algorithm.

      Keywords: Computer Vision, Real Time Image Processing, Surveillance, People Tracking, Visual Security System, Human Action Recognition.