[en] Millimeter-wave (mmWave) radar sensors have emerged as critical monitoring devices in various indoor applications, owing to their resilience to environmental conditions, non-intrusiveness, and cost-effectiveness. This thesis presents advancements in radar signal processing techniques tailored for indoor scenarios, with a focus on enhancing two radar output representations of a scene observed by distributed sensors: spatial reflectivity images and detection point clouds.
The thesis is divided into two parts, in the first part, it addresses challenges associated with generating high-quality reflectivity images from the reflected signals measured by widely distributed radar sensors. Leveraging compressed sensing methods, novel algorithms based on the Alternating Direction Method of Multipliers (ADMM) optimization framework are proposed to reconstruct global reflectivity images. Additionally, a heuristic method to accelerate the convergence of the proposed algorithms and reduce their computational complexity is introduced. Moreover, an efficient implementation of sparsity-based image reconstruction algorithms is proposed, achieved through automatic tuning of the regularization parameters while also considering synchronization errors.
The second part is devoted to the design of statistical detectors for mmWave radar sensors, focusing on detecting aspect-dependent targets and leveraging occlusion modeling in hypothesis testing formulations. Novel formulations are proposed, leading to 1) a detector with optimum weights on the processed signals of distributed sensors to jointly and efficiently detect aspect-dependent targets and 2) a detector based on occlusion modeling which enhances detection performance at each sensor node by leveraging the sparse structure of range profiles due to occlusions. Through scenario-based simulations and model-based evaluations, the proposed detectors demonstrate improved accuracy in detecting aspect-dependent and non-occluded targets, thereby providing accurate high-resolution point clouds.
The algorithms presented in this thesis provide an enhancement of the quality of radar images aiming to facilitate the subsequent analysis using classical image processing or advanced deep-learning techniques in various indoor applications.
Disciplines :
Electrical & electronics engineering
Author, co-author :
MURTADA, Ahmed Abdelnaser Elsayed ; University of Luxembourg > Interdisciplinary Centre for Security, Reliability and Trust (SNT) > SPARC
Language :
English
Title :
Radar Distributed Sensors for Indoor Imaging
Defense date :
10 June 2024
Institution :
Unilu - University of Luxembourg, Luxembourg
Degree :
Docteur en Informatique (DIP_DOC_0006_B)
Promotor :
MYSORE RAMA RAO, Bhavani Shankar ; University of Luxembourg > Interdisciplinary Centre for Security, Reliability and Trust (SNT) > SPARC
President :
AOUADA, Djamila ; University of Luxembourg > Interdisciplinary Centre for Security, Reliability and Trust (SNT) > CVI2
Jury member :
ALAEE, Mohammad ; University of Luxembourg > Interdisciplinary Centre for Security, Reliability and Trust (SNT) > SPARC
Schroeder, Udo; IEE S.A.
Diewald, Andreas; Trier University of Applied Sciences
Focus Area :
Computational Sciences
Development Goals :
9. Industry, innovation and infrastructure
FnR Project :
FNR15364040 - Radar Distributed Sensors For Indoor Imaging, 2020 (01/10/2020-15/06/2024) - Ahmed Abdelnaser Elsayed Murtada