Object Recognition in High-Resolution Indoor THz SAR Mapped Environment

Research output: Contribution to journalResearch articleContributedpeer-review

Contributors

  • Aman Batra - , University of Duisburg-Essen (Author)
  • Fawad Sheikh - , University of Duisburg-Essen (Author)
  • Maher Khaliel - , University of Duisburg-Essen, Benha University (Author)
  • Michael Wiemeler - , University of Duisburg-Essen (Author)
  • Diana Gohringer - , Chair of Adaptive Dynamic Systems (Author)
  • Thomas Kaiser - , University of Duisburg-Essen (Author)

Abstract

Synthetic aperture radar (SAR) at the terahertz (THz) spectrum has emerging short-range applications. In comparison to the microwave spectrum, the THz spectrum is limited in propagation range but benefits from high spatial resolution. The THz SAR is of significant interest for several applications which necessitate the mapping of indoor environments to support various endeavors such as rescue missions, map-assisted wireless communications, and household robotics. This paper addresses the augmentation of the high-resolution indoor mapped environment for object recognition, which includes detection, localization, and classification. Indoor object recognition is currently dominated by the usage of optical and infrared (IR) systems. However, it is not widely explored by radar technologies due to the limited spatial resolution at the most commonly used microwave frequencies. However, the THz spectrum provides a new paradigm of possible adaptation of object recognition in the radar domain by providing image quality in good compliance to optical/IR systems. In this paper, a multi-object indoor environment is foremost mapped at the THz spectrum ranging from 325 to 500 GHz in order to investigate the imaging in highly scattered environments and accordingly create a foundation for detection, localization, and classification. Furthermore, the extraction and clustering of features of the mapped environment are conducted for object detection and localization. Finally, the classification of detected objects is addressed with a supervised machine learning-based support vector machine (SVM) model.

Details

Original languageEnglish
Article number3762
JournalSensors
Volume22
Issue number10
Publication statusPublished - 15 May 2022
Peer-reviewedYes

External IDs

PubMed 35632171
ORCID /0000-0003-2571-8441/work/159607568

Keywords

Keywords

  • High-resolution indoor environment, Indoor imaging, Indoor object recognition, Synthetic aperture radar, Terahertz imaging