Perimeter-based calculation of object coverage in multi-camera systems
Résumé
For many visual monitoring scenarios, it may be desired to accurately compute the segment of an object’s perimeter that is being viewed by a particular camera, e.g. supporting applications for pattern recognition systems and coverage optimization. As objects may have different formats, generic polygon-based approaches may bring inaccurate results, since some simplifications may be taken. In this context, it is proposed in this paper an algorithm to accurately calculate the objects’ contours that can be viewed by a set of cameras, employing a discretization technique. The proposed algorithm computes the percentage of an object’s perimeter that is being viewed, which may be used as a parameter of coverage quality. Therefore, besides accurately computing effective viewing of cameras, a visual coverage quality metric for a set of objects is derived, which can be directly exploited for optimizations in camera surveillance systems and visual sensor networks.