Syntactic Pattern Recognition System Research Articles

Syntactic Simulation of 1-D Sedimentary Sequences in a Coal-Bearing Succession Using a Stochastic Context-free Grammar

ABSTRACT Syntactic pattern recognition (SPR) provides a new symbolic computing paradigm that has a great potential for application in sedimentologic and stratigraphic theory. In an SPR system, a formal language consists of a set of sentences; each sentence is a string of terminal symbols (or words); a grammar is used to characterize syntactic structures of the language, which tells whether or not a specific sentence belongs to the language. A language represents a class of patterns that are similar to each other in some sense. A sentence from the language then represents an individual pattern of the class. The grammar defines the legality of members of the class in general. There are several key subdomains in this methodology, such as grammar inference, language generation or sim lation, and language parsing. The objective of this paper is to show the possibility of using geological knowledge, obtained by domain experts (here sedimentologists), to construct valid grammars characterizing the vertical structure of different orders of sedimentary sequences, by applying the methodology to the analysis of a coal-bearing succession of the upper Limestone Coal Group, Pendleian (E1) in the Kincardine basin, Central Scotland. First, complete successions from two boreholes are encoded into syntactic strings of lithofacies symbols. Five groups of different high-frequency sedimentary sequences have been recognized. These high-frequency sequences combine to form a lower-frequency sequence. This geological knowledge is used in a heuristic sense to manually construct grammars f r each of the five groups of sequences and for the whole succession. Finally, the constructed grammars, together with a thickness simulation for each lithofacies, are used to generate similar sedimentary sequences to test the validity of the grammars and therefore their further application to a formal analysis of sedimentary sequences. Comparison of simulated sequences with real sequences indicates that SPR is an encouraging technique in the analysis of sedimentary sequences.

Pattern representations and syntactic classification of radar measurements of commercial aircraft

A syntactic pattern recognition system is evaluated for applications to radar signal identification. Three different level-crossing-based pattern representation algorithms are considered. The utility of the resulting symbolic pattern representations is assessed by evaluating the performance of a maximum-likelihood classifier when the observed symbol strings are used as inputs to the decision algorithm. A syntax analysis algorithm is derived from the likelihood function classifier. Performance results of simulated classification experiments for both maximum-likelihood and language-theoretic classifiers are presented.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

A Parallel Algorithm for Stochastic Image Segmentation

A parallel algorithm for syntactic image segmentation is introduced. Stochastic tree grammar is used as a context-generating model. It is shown that when this context-generating process is in the equilibrium state, a matched filter can be designed and applied in parallel to the image. This process can be used for image segmentation in a syntactic pattern recognition system to enhance the performance of the succeeding recognition process.

Inexact matching of line drawings in a syntactic pattern recognition system

A “picture pattern” is a description of a picture or class of pictures in terms of its components and their interrelationships. In the ESP 3 language and pattern recognition system, a line drawing is searched for occurrences of subpictures which match a given picture pattern. In “inexact-match mode”, the search routines are lenient in their interpretation of a pattern. Shapes of primitives need not be perfect, and predicates have built-in tolerance. Thus a simple pattern can be used to match a wide variety of noisy or distorted pictures. Inexact-match mode has been added to an experimental implementation of ESP 3. (13) The system has successfully matched a number of simple patterns such as alphabetic characters and has also located objects in some more complex line drawings such as a table and chairs scene.