Efficient camera selection strategy for multiple objects association

In this paper, we propose an efficient camera selection strategy finding the correspondence of objects among multiple cameras. Targets are associated by locally generating homographic lines. The effectiveness of homographic lines depends on the separation between targets, which is varied by different camera views. A pair of the best cameras are selected by using the amount of the separation between targets to minimize unnecessary homographic lines. The proposed method reduces computational costs for using homographic lines while achieving the maximum association performance. Association information in unselected cameras is updated with the constructed global information of objects in selected cameras. The proposed method is evaluated with the simulation and its computational costs are compared with the worst case.