Image Representation, Clustering, and Search in Proximity Graphs and Pathfinder Networks.
Abstract
This research extended Pathfinder networks and proximity graphs to new domains, and resulted in new proximity graphs. Pathfinder networks for both attributes and concepts can now be generated from the data used to construct a concept lattice (R. Wille), which can be represented by an overlay of the networks on the lattice. Growing sphere graphs (GSGs) model energy dispersion, and can generate sphere of influence graphs (SIGs), the union of mintrees (the sparsest Pathfinder network), or more general graphs, depending upon constraints. K-local image graphs (KLIGs) store information about the neighborhood surrounding each node in an arbitrary dynamic network. KLIGs provide a mechanism for planning, so that routing under conditions of failed nodes or edges can be near optimum. All minimum-cost paths between any pair of nodes in a KLIG consist of edges of the Pathfinder network PFN(r=1, q-n - 1). Pathfinder networks and other proximity graphs can now be generated dynamically by counting co-occurrences of events of interest, thus ensuring the incorporation of both clustering information and optimal paths through the graphs. A new paradigm for controlling weapons delivery vehicles (DVs) has been developed. Called Procrustes (it ensures that adequate resources are available), it provides a new way of viewing targets and DVs which utilizes a systematic way of substituting for failed DVs. It is a robust way of ensuring the success of such missions.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 15, 1996
- Accession Number
- ADA307726
Entities
People
- Donald W. Dearholt
Organizations
- Mississippi State University