| Article ID | Journal | Published Year | Pages | File Type |
|---|---|---|---|---|
| 532430 | Pattern Recognition | 2011 | 9 Pages |
The problem of locating visual sensors can be often modelled as 2D Art Gallery problems. In particular, tasks such as surveillance require observing the interior of a polygonal environment (interior covering, IC), while for inspection or image based rendering observing the boundary (edge covering, EC) is sufficient. Both problems are NP-hard, and no technique is known for transforming one problem into the other. Recently, an incremental algorithm for EC has been proposed, and its near-optimality has been demonstrated experimentally. In this paper we show that, with some modification, the algorithm is nearly optimal also for IC. The algorithm has been implemented and tested over several hundreds of random polygons with and without holes. The cardinality of the solutions provided is very near to, or coincident with, a polygon specific lower bound, and then suboptimal or optimal. In addition, our algorithm has been compared, for all the test polygons, with recent heuristic sensor location algorithms. In all cases, the cardinality of the set of guards provided by our algorithm was less than or equal to that of the set computed by the other algorithms. An enhanced version of the algorithm, also taking into account range and incidence constraints, has also been implemented.
