Affiliations: [a] Dipartimento di Scienze dell'Ingegneria, Università di Modena, Via Campi 213/b, 41100 Modena, Italy. E-mail: [email protected] | [b] D.E.I.S., Università di Bologna, V.le Risorgimento 2, 40136 Bologna, Italy. E-mail: [email protected] | [c] Dipartimento di Ingegneria, Università di Ferrara, Via G. Saragat 1, 44100 Ferrara, Italy. E-mail: [email protected], [email protected]
Abstract: We present an application of machine learning and statistics to the problem of distinguishing between defective and non-defective industrial workpieces, where the defect takes the form of a long and thin crack on the surface of the piece. From the images of pieces a number of features are extracted by using the Hough transform and the Correlated Hough transform. Two datasets are considered, one containing only features related to the Hough transform and the other containing also features related to the Correlated Hough transform. On these datasets we have compared six different learning algorithms: an attribute-value learner, C4.5, a backpropagation neural network, NeuralWorks Predict, a k-nearest neighbour algorithm, and three statistical techniques, linear, logistic and quadratic discriminant. The experiments show that C4.5 performs best for both feature sets and gives an average accuracy of 93.3% for the first dataset and 95.9% for the second dataset.
