MACHINE VISION FOR COMPOSITE MANUFACTURING

MANUFACTURE OF LAMINATED COMPOSITE MATERIAL IS A LABOR INTENSIVE AND THUS AN EXPENSIVE PROCESS. A CRITICAL PROBLEMIS THE INSPECTION OF PLY EDGES TO DETERMINE WHETHER A PLY EDGE HAS BEEN PLACED WITHIN TOLERANCE. FLOW HAS DEVELOPED AMACHINE VISION SYSTEM TO AUTOMATE THIS INSPECTION PROCESS. THIS PROPOSED RESEARCH EFFORT SEEKS TO IMPROVE THE ACCURACY,ADAPTABILITY, AND SPEED OF THIS SYSTEM BY THE USE OF NEURAL NETWORK TECHNIQUES. THE PROPOSED IMAGE PROCESSING SYSTEM WILL USE A REGULAR ARCHITECTURE THAT INTEGRATES NEURAL NETWORK MODULES IN LAYERS. IT IS EXPECTED THAT CONSTRAININGTHE SYSTEM TO A HIGHLY REGULAR ARCHITECTURE CAN SIGNIFICANTLY INCREASE THE LEARNING SPEEDS FOR NEURAL NETWORKS. EACH NETWORK MODULE WILL BE RELATIVELY SIMPLE, AND THERE WILL BE ONLY A FEW DIFFERENT TYPES OF MODULES. THE SYSTEM CAN BE MANUFACTURED USING EXISTING VLSI TECHNOLOGY. EXPECTED ADVANTAGES FROM THIS APPROACH ARE: DATA COMPRESSION, INCREASED SPEED, CAPABILITY OF WORKING WITH AMBIGUOUS IMAGES, CAPABILITY OF HANDLING A LARGE FIELD OF VIEW WITH GOOD RESOLUTION, AND REDUCED COST.