Glass Crack Defect Detection using BPSO and Support Vector Machine Classifier

Abstract

The paper deals with the optimization of an artificial vision system for the detection of defects in the glass. This system is integrated in the middle of the production chain of these glasses to minimize the arrival of defective glasses to our customers.

With the optimization of the artificial vision system, we intend to achieve a higher performance of the system. In order to make these improvements, we have to carry out a study on the behaviour of the system, once it is integrated into our production line, since the conditions of each company vary and make the system perform worse or better.

In order to understand the possible problems that we may encounter with glass analysis, it was important to include a brief explanation of the characteristics of the glass, the production process and the operation of the artificial vision system followed by the feature reduction using binary particle swarm optimization (BPSO) and classification using support vector machine (SVM) classifier. The simulation results is expressed in terms of a confusion matrix plot with precision, recall, accuracy and F-score values.