Auto-Determination of Camera Parameters for Scenario Simulations in Visual Surveillance Applications

We propose a novel method for determining geometric parameters of general cameras to generate augmented reality images. Visual surveillance systems are usually evaluated by using datasets from different camera configurations. Then, the system is often fixed and one cannot ensure that the same methods can work in all configurations. Our approach deals with the autodetermination of geometric parameters from the images to generate virtual views situations, and evaluate quantitatively visual surveillance systems in any configuration. The method is based on geometry-free characterization of the images using a common codebook and random ferns to find correspondence points. Experimental results are presented on real data from two different locations for people collapse scenarios.