Compressed Sensing in Material Diagnostics via Ultrasound Imaging (CoSMaDU)

• Applicants: Professor Dr.-Ing. Giovanni Del Galdo; Professor Dr.-Ing. Ahmad Osman
• Subject Area: Electronic Semiconductors, Components and Circuits, Integrated Systems, Sensor Technology, Theoretical Electrical Engineering ; Measurement Systems
• Term: from 2019 to 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 421389590

• Final Report Year: 2024

Final Report Abstract

In the Compressed Sensing in Material Diagnostics via Ultrasound imaging (CoSMaDU) project, we have addressed the challenges of ultrasound modelling and high resource demands (e.g. number of sensors, increasing data rates) by investigating linear models that more adequately capture ultrasound propagation while still lending themselves to being incorporated into the powerful algorithmic framework of signal processing. In particular, research was conducted in the direction of Compressive Sensing (CS) at the algorithmic and measurement architecture levels. As part of our results, we have developed new reconstruction frameworks integrating enhanced forward modelling techniques, namely series expansions and ray casting for complex geometries and multiple scattering and CS. At the same time, the compression schemes were designed in such a way that they are not only interesting from an academic standpoint, but can also lend themselves to realistically achievable hardware implementations. This was achieved by considering compression through spatial and frequency subsampling, for which optimal subsampling patterns were studied. The obtained patterns provide gains regarding measurement speed, reconstruction accuracy, or in some cases both. In the case of handheld measurements, naturally occurring spatial subsampling was also studied. The forward modelling and subsampling approaches obtained throughout this project were incorporated into CS reconstruction frameworks and applied to pipe inspection and computed tomography scenarios, showing good results. These results were combined with Machine Learning (ML), showing promising results and high affinity with current topics such as model-based Deep Learning (DL) and adaptive sensing. In addition, theoretical performance bounds were employed to show both the performance of the subsampling patterns and the impact of accurate modelling when reconstructing defect maps. The CoSMaDU contributions amount to a total of 11 publications, as well as a number of bachelor and master theses and an additional paper submitted to a journal.

Publications

• 3D reconstruction of handheld data by SAFT and the influence of measurement inaccuracies. 2019 IEEE International Ultrasonics Symposium (IUS), 2095-2098. IEEE.
  Krieg, Fabian; Kodera, Sayako; Kirchhof, Jan; Romer, Florian; Ihlow, Alexander; Lugin, Sergey; Osman, Ahmad & Galdo, Giovanni Del
  
• Total Focusing Method with Subsampling in Space and Frequency Domain for Ultrasound NDT. 2019 IEEE International Ultrasonics Symposium (IUS), 2103-2106. IEEE.
  Perez, Eduardo; Kirchhof, Jan; Semper, Sebastian; Krieg, Fabian & Romer, Florian
  
• Cramér-Rao Bounds for Flaw Localization in Subsampled Multistatic Multichannel Ultrasound Ndt Data. ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 4960-4964. IEEE.
  Perez, Eduardo; Kirchhof, Jan; Semper, Sebastian; Krieg, Fabian & Romer, Florian
  
• Subsampling Approaches for Compressed Sensing with Ultrasound Arrays in Non-Destructive Testing. Sensors, 20(23), 6734.
  Pérez, Eduardo; Kirchhof, Jan; Krieg, Fabian & Römer, Florian
  
• Compressed Sensing: From Big Data to Relevant Data. Handbook of Nondestructive Evaluation 4.0, 1-24. Springer International Publishing.
  Römer, Florian; Kirchhof, Jan; Krieg, Fabian & Pérez, Eduardo
  
• Compressed Ultrasound Computed Tomography in NDT. 2021 IEEE International Ultrasonics Symposium (IUS), 1-4. IEEE.
  Perez, Eduardo; Semper, Sebastian; Kirchhof, Jan; Krieg, Fabian & Romer, Florian
  
• Frequency Subsampling of Ultrasound Nondestructive Measurements: Acquisition, Reconstruction, and Performance. IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, 68(10), 3174-3191.
  Kirchhof, Jan; Semper, Sebastian; Wagner, Christoph W.; Perez, Eduardo; Romer, Florian & Del, Galdo Giovanni
  
• Locally Optimal Subsampling Strategies for Full Matrix Capture Measurements in Pipe Inspection. Applied Sciences, 11(9), 4291.
  Krieg, Fabian; Kirchhof, Jan; Pérez, Eduardo; Schwender, Thomas; Römer, Florian & Osman, Ahmad
  
• Preprocessing of Freehand Ultrasound Synthetic Aperture Measurements using DNN. 2021 29th European Signal Processing Conference (EUSIPCO), 1401-1405. IEEE.
  Pandey, Rick; Kirchhof, Jan; Krieg, Fabian; Perez, Eduardo & Romer, Florian
  
• Deep Learning Aided Interpolation of Spatio-Temporal Nonstationary Data. 2022 30th European Signal Processing Conference (EUSIPCO), 2221-2225. IEEE.
  Kodera, Sayako; Romer, Florian; Perez, Eduardo; Kirchhof, Jan & Krieg, Fabian
  
• Deep Learning-Based Optimal Spatial Subsampling in Ultrasound Nondestructive Testing. 2023 31st European Signal Processing Conference (EUSIPCO), 1863-1867. IEEE.
  Wang, Han; Pérez, Eduardo & Römer, Florian