Sequential geometrical pattern recognition

The primary goal of this research was to contribute to the technological advancement of computer aided design in the areas of man/computer interface, data compression, information storage and retrieval, and machine generation of engineering drawings. A key factor in realizing this overall objective was the development and implementation of a unique pattern recognition algorithm for the purpose of efficiently and accurately identifying a defined set of unique electronic symbols and circuit interconnections. The algorithm developed utilizes physical characteristics of a digitized image to sequentially reclassify the pattern image into smaller areas of classification until it is uniquely identifiable. It contains error detection and correction logic designed to minimize the effect of generated noise and data drop out. It is also modular in design and can be modified with relative ease to provide for future data set expansion. To provide the data base for evaluating the algorithm and effectiveness of the overall approach, sixty unique electronic symbols and circuit interconnections were established as a data set. The evaluation consisted of using the recognition algorithm to uniquely identify specific symbols and interconnections, storing the compressed data on magnetic tape, and subsequently generating the specified engineering drawings from the magnetic tape using an X-Y Plotter. The full potential of the approach for improving the man/computer interface was not demonstrated because of the unavailability of a video scanner and digitizer. However, the feasibility of the technique was established by simulating the scanner/digitizer output with binary coded card deck images. This alternate approach of inputting computer data provided an effective method for controlling pattern image error generation during the development and evaluation of error detection and correction logic. Error detection and correction logic was developed and implemented to perform majority voting during the recognition of primary features. Data compression, achieved through the pattern recognition and coding process, resulted in efficient information storage and retrieval both in dedicated memory storage and access time. Four test cases were used to evaluate the accuracy of the recognition algorithm and the effectiveness of the approach used in data compression, storage and retrieval, and circuit reconstruction. Three electronic circuit schematic diagrams were accurately constructed from a pool of unidentified pattern images. The fourth test case consisted of recognizing, compressing, storing, retrieving and constructing a symbol matrix containing the entire set of sixty symbols and circuit interconnections in the specified sequence.