LARGE NUMBERS OF INDUSTRIAL RADIOGRAPHS ARE GENERATED TO INSPECT HIGH RELIABILITY ITEMS SUCH AS PIPE WELDS, BOILER SEAMS AND CASTINGS.

LARGE NUMBERS OF INDUSTRIAL RADIOGRAPHS ARE GENERATED TO INSPECT HIGH RELIABILITY ITEMS SUCH AS PIPE WELDS, BOILER SEAMS AND CASTINGS. THE INTERPRETATION OF THESE RADIOGRAPHSCAN BE DIFFICULT AND TIME-CONSUMING BECAUSE OF LOW IMAGE CONTRAST, NOISE, EXPOSURE VARIABILITY, EXPOSURE NON-UNIFORMITY AND OTHER EXPERIMENTAL FACTORS. IT IS THE PURPOSE OF THE PROPOSED RESEARCH TO DEMONSTRATE THE FEASIBILITY OF COMPUTER AIDED RADIOGRAPH INTERPRETATION. THE WORK WILL INCLUDE ALGORITHMS TO GENERATE QUANTITATIVE MEASURES OF RADIOGRAPH QUALITY, AN AREA SELECTIVE NON-LINEARCONTRAST ENHANCEMENT SCHEME TO MAKE DEFECTS STAND OUT, AND COMPUTER PRESCREENING OF RADIOGRAPHS TO LOCATE POTENTIAL DEFECTS FOR MORE CAREFUL EXAMINATION BY RADIOGRAPH INTERPRETERS. HARDWARE CONCEPTS FOR MULTIPLE CAMERA, HIGH RESOLUTION, RAPID RADIOGRAPH DIGITIZATION WILL ALSO BE DEMONSTRATED. THE WORK WILL USE A STATE-OF-THE-ART INDUSTRIAL COMPUTER VISION SYSTEM AS A TEST BED.