Treffer: Automated flaw detection in X-ray tomography of polymer insulators.

In power distribution networks, polymer insulators commonly show flaws in their interior which cannot be perceived without the use of a non-destructive inspection mechanism. This work describes a methodology based on three-dimensional industrial computer tomography (CT) for the automatic detection of flaws in polymer insulators. CT slices were reconstructed using 180 digital radiographs (projections) acquired by a high-resolution system (pixel dimension of 50 μm × 50 μm). For the reconstruction of 3D CT, 50 μm-wide bitmap slices were used. 3D segmentation and reconstruction is performed by the Marching Cubes algorithm implemented in the Visualization Toolkit (VTK) open-source library and in the Java programming language. The final step of the detection mechanism is performed by a neural network trained using nine features obtained from the reconstructed 3D objects. The estimated accuracy obtained by the proposed techniques is lower than that of the conventional technique. However, the available testing data are scarce and this makes it difficult to derive a definite conclusion on the correctness rate. The most important scientific contribution of this work resides in the fact that it is a promising work which may have great potential as far as automatic detection of flaws in polymer insulators based exclusively on the 3D domain is concerned, but this analysis must be addressed in future work. [ABSTRACT FROM AUTHOR]