Trajectory retrieval is the process of obtaining the path of a moving object from large set of data of videos. Allowing the users to be able to obtain these paths from the databases of videos containing an enormous number of trajectories, helps in speeding up the searching process for a specific activity or a target object, instead of searching it manually by observing each video with each detail to get the object in concern. This system is applied mainly for video surveillance applications in order to ease the process of exploring or detecting any suspicious activity or dangerous events. The system is based on an interface that is constructed to allow and facilitate the interaction between the user and the database trajectories. In addition, this interface represents the results of the retrieved trajectories by a visual feedback.

This thesis will discuss how the system’s goal can be achieved through introducing a similarity measure that is capable of identifying the trajectories in large data sets that are similar to that entered by the user. An approach for the similarity measure is applied which is the global alignment technique for matching between the trajectories. The results will show that similar trajectories can be produced from the database through analyzing and classifying the trajectories according to the similarity of their shape. Therefore will help the user in minimizing and reducing the paths of objects to be analyzed and monitored.