Developing Parallel Cracks and Spots Ceramic Defect Detection and Classification Algorithm Using CUDA

Abstract

Automatic defects inspection and classification inceramic tile plays a crucial role in ceramic tiles industry. Inorder to improve the required computing time to detect andclassify spot and crack defects in ceramic tiles, this paperproposed parallel algorithms based on the graphical processingunit. The proposed algorithm divides ceramic tile images intonon-overlapped partitions, identifies the defected partitions andthen carries out the defect classification algorithm only onthe defected partitions. Simulation results indicate significantimprovements in terms of classification time in comparisonto the current techniques. Furthermore, computation time forparallel algorithm excels the sequential algorithm. The resultsshow the speed up of the computation of the proposed parallelcompared with sentential algorithms that increase as the imagesize increases. Finally, the speed up of 45 times has been accomplished.