A SYSTEM FOR MACHINE READING OF CHEMICAL SYMBOLS

THIS PROJECT WILL LEAD TO THE DEVELOPMENT OF A PROOF-OF-PRINCIPLE SYSTEM FOR MACHINE READING AND UNDERSTANDING OF HARDCOPY, GRAPHICAL CHEMICAL STRUCTURES. THIS RESEARCH WILL PROVIDE A BASIS FOR DEVELOPING INTELLIGENT INPUT DEVICES FOR ENTRY OF CHEMICAL STRUCTURE DATA INTO EXISTING AND FUTURE CHEMICAL INFORMATION ANALYSIS AND CLASSIFICATION SYSTEMS. SUCH SYSTEMS HAVE APPLICATIONS FOR CHEMICAL PATENT SEARCHES, CHEMICAL DESIGN AND ANALYSIS, AND EDUCATION. DEVELOPMENT OF THE PROOF-OF-PRINCIPLE CHEMICAL STRUCTURE READER WILL REQUIRE A UNIQUE BLEND OF CURRENT TECHNOLOGIES, INCLUDING ELECTRONIC IMAGING, PATTERN RECOGNITION AND ANALYSIS, AND COMPUTER HANDLING OF CHEMICAL INFORMATION. THE READER WILL BE A MICROCOMPUTER-BASED SYSTEM USING A CCD CAMERA TO CAPTURE PAGE IMAGES CONSISTING OF CHEMICAL STRUCTURE DIAGRAMS. THESE IMAGES WILL BE DIGITIZED AND TRANSMITTED TO THE COMPUTER WHERE THE IMAGE DATA WILL BE PREPROCESSED INTO A FORMAT UNDERSTANDABLEBY PATTERN RECOGNITION ALGORITHMS. THE ALGORITHMS WILL TRANSFORM THE DATA FROM A COLLECTION OF GRAPHIC ELEMENTS INTO CHEMICAL STRUCTURE INDICES CHARACTERIZING THE CHEMICAL.THE SPECIFIC GOAL OF THE PHASE I EFFORT WILL BE TO DEVELOP ASYSTEM CAPABLE OF AUTOMATICALLY READING CHEMICAL STRUCTURE DIAGRAMS FROM A LIMITED SET AND CORRECTLY IDENTIFYING THE CHEMICAL.