Simulator Driven Machine Learning on Embedded Systems for GPR Navigation and IED Detection, Akronym: MEDICI-LIBERTAD

• Applicants: Dr.-Ing. Christoph Baer; Professor Dr.-Ing. Michael Hübner; Dr.-Ing. Christian Schulz
• Subject Area: Electronic Semiconductors, Components and Circuits, Integrated Systems, Sensor Technology, Theoretical Electrical Engineering
• Term: from 2019 to 2023
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 423771041

• Final Report Year: 2023

Final Report Abstract

Several strategies for AI-based sensor evaluation in the context of radar-based humanitarian demining were successfully examined in project MEDICI-LIBERTAD. These studies comprised an unique radar-based locating system, an optimized 2D EM simulator, and AI application in embedded devices. The 2D simulator can easily generate synthetic and randomized GPR data for AI-based detection algorithms that are supervised and unsupervised. Its viability has been demonstrated in well-known detection techniques such as YOLO and Detectron 2, as well as in algorithms used in embedded systems. Regarding the radar-based positioning system, a novel idea called channel fingerprinting was researched and presented. It enables the calculation of a highly precise position in 2D settings using only a single radar measurement. Based on the simulator, the effect of misplacements is cross-checked and suitable thresholds derived. All of the project achievements that were attained advance the state of the art in general, and demining technology in particular. Future research initiatives will make use of the project outcomes. Additionally, the obtained results can be adapted to other fields of research and crossfertilize investigations in the field of radar positioning and highly optimized / multi-physical simulations. Because partners from Universidad Nacional de Colombia actively participated in the project, it is important to note that MEDICI-LIBERTAD was carried out as an international project. Consequently, it increased the universities' international contacts and allowed students to undertake part of their theses abroad on global research issues, which is highly gratifying in addition to the pure research work.

Publications

• A Novel Localization System in SAR-Demining Applications Using Invariant Radar Channel Fingerprints. Sensors, 22(22), 8688.
  Karsch, Nicholas; Schulte, Hendrik; Musch, Thomas & Baer, Christoph
  
• A Statistical FDFD Simulator for the Generation of Labeled Training Data Sets in the Context of Humanitarian Demining using GPR. 2022 IEEE MTT-S International Conference on Numerical Electromagnetic and Multiphysics Modeling and Optimization (NEMO) (2022, 7, 6), 1-3. American Geophysical Union (AGU).
  Schorlemer, Jonas; Jebramcik, Jochen; Baer, Christoph; Rolfes, Ilona & Schulz, Christian
  
• Cognitive FMCW-Radar Concept for Ultrafast Spatial Mapping using Frequency Coded Channels. 2021 18th European Radar Conference (EuRAD) (2022, 4, 5). American Geophysical Union (AGU).
  Karsch, Nicholas; Baer, Christoph & Musch, Thomas
  
• Comparison of Short-Range SAR Imaging Algorithms for the Detection of Landmines using Numerical Simulations. 2021 18th European Radar Conference (EuRAD) (2022, 4, 5). American Geophysical Union (AGU).
  Schorlemer, Jonas; Jebramcik, Jochen; Rolfes, Ilona & Barowski, Jan
  
• Detecting Improvised Land-mines using Deep Neural Networks on GPR Image Dataset targeting FPGAs. 2022 IEEE Nordic Circuits and Systems Conference (NorCAS) (2022, 10, 25), 1-7. American Geophysical Union (AGU).
  Mahmood, Safdar; Scharoba, Stefan; Schorlemer, Jonas; Schulz, Christian; Hubner, Michael & Reichenbach, Marc