Electrotechnical parts quality control by computed tomography

This paper deals with the quality assessment of electrotechnical parts by means of computed tomographic images. The quality evaluation of such elements is achieved nowadays by experts carrying out experiments but this task is time consuming. The proposed approach is based on the computation of several attributes to characterize the sought-after regions. Four methods are applied to measure the local organisation of the glass fibers in the element: the region growing algorithm, the principal component analysis of gradient vectors, the wavelet transform and the Radon transform. Each of them are images which bring complementary and redundant information on the electrotechnical part internal organisation. Then, in a second step, the attributes are aggregated by a fusion system based on simple rules in order to obtain a global segmentation of the computed tomographic images.