Direction-based vacancy queries in camera sensor networks

In a camera sensor network, given a target q and a set of cameras \({\mathcal P}\), q may not be recognized by \({\mathcal P}\) due to its facing direction even if every \(p_{i} \in {\mathcal P}\) can see q. In this paper, we study the direction-based vacancy query, which can find out the vacant direction ranges where q cannot be recognized by any cameras. If q’s facing direction \(\mathbf {f_{q}}\) is far from the sensing direction of a camera, i.e., the included angle between the two directions is smaller than \(\theta \), q cannot be effectively recognized due to the vacancy of cameras on the direction \(\mathbf {f_{q}}\). To answer vacancy queries, the basic algorithm is to sort the directions of all cameras \({\mathcal P}\) and then to identify the vacant ranges. To make the vacancy query faster, we design an index structure, i.e., \(\alpha \)-polygon tree, which can organize the objects according to their directions. To make the tree as balance as possible, we propose an algorithm to select the best splitter for each tree node when building the tree. Our tree-based algorithm is to shrink the vacancy ranges while visiting the tree nodes in a depth-first order. The tree nodes are pruned if they do not contain objects (i.e., cameras) that may make the vacancy ranges shrink. We conducted experiments to evaluate the performances of the \(\alpha \)-polygon tree and the tree-based query algorithm on both synthetic and real datasets.