Efficient analysis of high-dimensional ultrasound data in non-destructive testing
Joint BMBF research project
ZeMat4: Efficient algorithms for defect classifictation

Project coordinator: Prof. Dr. Armin Iske
Project member: José Fernando Cuenca Jimenez

Classification and quantification plays an important role in non-destructive testing. Hence, effective numerical analysing methods are needed. However, the underlying data is not only huge and heterogeneous but also high dimensional and therefore highly complex. Thus, the use of powerful dimension reduction methods is required. In this project non-linear projection methods based on modern geometrical concepts of machine learning will be applied. Thereby, great importance will be attached to geometrical, topological and problem specific invariance. An adaption and evaluation of the algorithms will be developed in cooperation with other subprojects.