Partial Decision Aggregation for Electrotechnical Parts Quality Control

3D tomographic images are used by the Schneider Electric Company to assess the quality of composite materials. These images bring information on the internal organization of the parts. Several image classification methods were applied to detect the orientated and the isotropic regions within the image. This paper presents the fusion of these partial classifications by means of the decision templates method proposed by L. Kuncheva. An adaptation of the initial method is proposed to manage the class of reject for which it is not possible to learn a decision template. Obtained detections are assessed by the confusion matrix whose results show the effectiveness of the approach