Developing a machine vision system for spangle classification using image processing and artificial neural network

These studies are carried out to classify the three different spangle patterns found on the galvanized steel sheets by image processing and artificial neural network. Images of 200 × 200 pixel sizes from three different spangle samples were captured using optical filter and digital camera. These images were preprocessed and Haralicks (energy, entropy, contrast and homogeneity) and Laws (LE/EL, LS/SL, LR/RL, ES/SE and SR/RS) textural parameters were calculated. Principle component analysis was carried out on the generated textural database and this database was used to train and test the artificial neural network. The artificial neural network could be able to classify the spangle pattern up to a reliable extent and the overall accuracy was 80.09% for investigated samples. The proposed methodology can be used for quantification of spangle patterns and to develop an online system for spangle classification. Matlab® 7 was used for image processing and artificial neural network studies.