Fuzzy Machine Vision Based Porosity Detection

In this paper an objective fuzzy approach for fast and accurate porosity vision based inspection is presented. An automated methodology of detection of pores, which are formed in aluminum alloys during production of water-pumps for car engines with the die casting method, is described. The proposed method is based on the correlation of the core of pore candidates with twelve developed matrices resulted in five novel features. The fuzzy decision making on porosity detection adopted and presented in this paper, adds great value to the whole production system, by increasing the confidence of the inspectors in the machine performing real-time verification. The fuzzy porosity detection was carried out on a database of 105 gray level images. The proposed model properly identifies 93.36% of the pores in the entire database.