Symbolic pattern processing system for planar spatial structures
We describe a symbolic pattern processing system that recognizes when composite planar structures are present in a two-dimensional spatial pattern. The composite objects are composed and identified by the arrangement of elementary objects (such as rectangles, squares, circles, etc.). The composite planar structures are represented by a special class of graphs where the nodes satisfy a two-dimensional ordering. We use straightforward algorithms to generate these graphs from an image (obtained using a vision system) that was preprocessed by a system of low-level neural networks. The graphs are generated for all arrangements of elementary objects in an image, and the resulting graphs are compared with predetermined models of composite objects in order to determine if the composite object is present in the image. We present an algorithm to find out if two given graphs match within a specified tolerance given by the pair (l,d). We then develop a similarity measure between two graphs based on two different linearizations of two-dimensional lattice points. A decision making module, based on empirical rules, decides if the image contains any instances of the prestored models of composite objects. 10 refs., 4 figs.
- Research Organization:
- Oak Ridge National Lab. (ORNL), Oak Ridge, TN (United States)
- Sponsoring Organization:
- DOE/ER
- DOE Contract Number:
- AC05-84OR21400
- OSTI ID:
- 7020403
- Report Number(s):
- CONF-900652-1; ON: DE90001598
- Resource Relation:
- Conference: 10. international conference on pattern recognition, Atlantic City, NJ (USA), 17-21 Jun 1990
- Country of Publication:
- United States
- Language:
- English
Similar Records
Digital and optical shape representation and pattern recognition; Proceedings of the Meeting, Orlando, FL, Apr. 4-6, 1988
Hybrid pattern recognition system capable of self-modification