Les travaux menés lors de cette thèse s’inscrivent dans le cadre général de la détection de cibles en utilisant une forme d’onde non-conventionnelle large bande. L’utilisation d’une forme d’onde large bande à faible PRF a été proposée par le passé une alternative aux traitements multi-PRF qui limitent le temps d’illumination de la scène. En effet, l’augmentation de la bande instantanée permet d’obtenir une meilleure résolution distance ; les cibles rapides sont alors susceptibles de migrer lors du temps de traitement, mais ce phénomène de couplage distance-vitesse peut être mis à profit pour lever les ambiguïtés. L’objectif de la thèse est alors de développer, pour une forme d’onde large bande avec faible PRF, des traitements prenant en compte la migration des cibles et capables de lever les ambiguïtés vitesse dans des scénarios réalistes. Les travaux se basent sur un algorithme de représentation parcimonieuse non-ambigüe de cibles migrantes, dans un cadre algorithmique Bayésien. Cet algorithme est en revanche développé sous certaines hypothèses, et des travaux de robustification sont alors entrepris afin de l’utiliser sur des scénarios plus réalistes. Dans un premier temps, l’algorithme est robustifié au désalignement des cibles par rapport à la grille d’analyse, puis modifié pour prendre également en compte une possible composante diffuse de bruit. Il est également remanié pour estimer correctement une scène comportant une forte diversité de puissance, où des cibles fortes masquent potentiellement des cibles faibles. Les différents algorithmes sont validés à la fois sur des données synthétiques et expérimentales. / The work conducted during this PhD falls within the general context of radar target detection using a non-conventional wideband waveform. More precisely, the use of a low-PRF wideband waveform has been proposed in the past as an alternative to the classical staggered-PRF processing used to mitigate velocity ambiguities that limits dwell time. Increasing the instantaneous bandwidth improves range resolution; fast moving targets are then likely to migrate during the coherent processing interval. This range-velocity coupling can then be used to mitigate velocity ambiguities. This PhD thesis aims at developing an algorithm able to provide unambiguous estimation of migrating targets using a low-PRF wideband waveform. It is based on a sparse representation algorithm able to unambiguously estimate migrating targets, within a Bayesian framework. However, this algorithm is developed under some hypothesis, and then requires robustification to be used on more realistic scenarii. First, the algorithm is robustified to the case of off-grid targets, and then upgraded to take into account a possible diffuse clutter component. On the other hand, the reference algorithm is modified to accurately estimate high dynamic range scenes where weak targets compete with strong targets. All the developed algorithms have been validated on synthetic and experimental data recorded by the PARSAX radar from the Technical University of Delft, The Netherlands.

Microwave imaging finds numerous applications involving optically obscured targets. One particular area is breast cancer detection, since microwave technology promises fast low-cost image reconstruction without the use of harmful radiation typical of X-ray mammography. However, the success of microwave imaging is hindered by a critical issue, the complex nature of near-field electromagnetic scattering in tissue. To overcome this, specialized image reconstruction algorithms alongside sensitive measurement hardware are required. In this work, real-time near-field microwave imaging algorithms known as quantitative microwave holography and scattered power mapping are explored. They are experimentally demonstrated to identify potential tumor regions in tissue phantoms. Alongside this development, quality control techniques for evaluating microwave hardware are also described. Two new methods for improving the image reconstruction quality are also presented. First, a novel technique, which combines two commonly used mathematical approximations of scattering (the Born and Rytov approximations), is demonstrated yielding improved image reconstructions due to the complimentary nature of the approximations. Second, a range migration algorithm is introduced which enables near-field refocusing of a point-spread function (PSF), which is critical for algorithms that rely on measured PSFs to perform image reconstruction. / Thesis / Doctor of Philosophy (PhD) / Breast cancer remains as one of the highest causes of cancer-related deaths in women in Canada. Though X-ray mammography remains the gold standard for regular breast cancer screening, its use of harmful radiation, painful breast compression, and radiologist dependent evaluation remain as detracting factors for its use. Over the past 40 years, researchers have been exploring the use of microwave technology in place of X-ray mammography. Microwave radiation, used at power levels similar to that of a cellphone, has been demonstrated successfully in simulations of breast scans. However, in experimental evaluations with breast phantoms, the complex scattering path of the radiation through tissue complicates image reconstruction. In this thesis, methods of improving the accuracy of microwave algorithms are explored, alongside new breast phantom structures that replicate well the electrical properties of tissue. The results of this work demonstrate the flexibility of microwave imaging, and the adversities that still need to be overcome for it to begin seeing clinical use.