Conference Paper

Brief Announcement: Pan and Scan

DOI: 10.1145/1835698.1835729 Conference: Proceedings of the 29th Annual ACM Symposium on Principles of Distributed Computing, PODC 2010, Zurich, Switzerland, July 25-28, 2010
Source: DBLP


We introduce the pan and scan problem, in which cameras are configured to observe multiple target locations. A camera's configuration consists of its orientation and its zoom factor or field or view (its position is given); the quality of a target's reading by a camera depends (inversely) on both the distance and field of view. After briefly discussing an easy setting in which a target accumulates measurement quality from all cameras observing it, we move on to a more challenging setting in which for each target only the best measurement of it is counted, for which we give various results. Although both variants admit continuous solutions, we observe that we may restrict our attention to solutions based on pinned cones. For a geometrically constrained setting, we give an optimal dynamic programming algorithm. For the unconstrained setting of this problem, we prove NP-hardness, present efficient centralized and distributed 2-approximation algorithms, and observe that a PTAS exists under certain assumptions. For a synchronized distributed setting, we give a 2-approximation protocol and a (2β)/(1-α)-approximation protocol (for all 0 ≤ α ≤ 1 and β ≥ 1) with the stability feature that no target's camera assignment changes more than logβ(m/α) times. We also discuss the running times of the algorithms and study the speed-ups that are possible in certain situations.

2 Reads