Database collection for performance analysis of object and event detection in surveillance videos

Abstract

In our era, surveillance systems are largely employed in the field of security and data gathering. The main driving force behind the expansion of these visual surveillance systems is due to the active use of visual surveillance systems in security applications. For these systems to be developed and to be able to act in real time, the performance of object and event detection algorithms must be improved. The objective comparison of detection algorithms will provide a concrete base to carry out research on this topic and lead to the measurement of real developments. To be able to conduct objective comparisons, databases which are accessible for research purposes and evaluation metrics are needed. Within the literature itself, a few evaluation metrics are defined, but databases that are accessible for research purposes are not common. This paper presents an analysis of a database which was formed within Sabancinodot University, based on surveillance systemspsila use of object and detection algorithms. Finally the performance analysis of an object detection algorithm that was tested on the database is presented.