A tree automaton system of syntactic pattern recognition for the recognition of seismic patterns

In a seismogram, certain structural seismic patterns always exist. So syntactic (structural) pattern recognition is proposed to recognize seismic patterns. In this study, error-correcting tree automata techniques of syntactic pattern recognition are proposed to recognize seismic patterns in the seismogram. A block diagram of a tree automaton system is presented. The system includes two parts. The training part transforms training patterns into tree grammar. The tree grammar is expandable by using grammatical inference and learning procedure on the tree grammar of training patterns. Error-correcting tree automata of syntax analysis are constructed from tree grammar. The recognition part of the system analyzes input seismograms and generates recognition results. Initially, the input seismogram is skeletonized by thinning algorithms, followed by pattern segmentation, primitive recognition, and tree representation construction procedures. Each seismic pattern is represented by a tree representation. Finally, error-correcting tree automata can recognize each input seismic pattern. The analyzed 2-D synthetic seismic patterns are bright spot, pinch-out, flat spot, gradual sea level fall, and gradual sea level rise patterns. From the experimental results, the tree approach for the recognition of seismic patterns is very encouraging and very important in automatic seismic pattern interpretation.