Surface Reconstruction using Multipath Propagation in Indoor MIMO Radar

Sammanfattning: Using active radar for localization in indoor environments is challenging. By signal reflection off walls and other static objects, false targets occur, referred to as multipath "ghosts". These are a significant hinder in interpreting the data. For this reason mapping out problematic surfaces that produce reflections is an important task, preferably without prior knowledge of the room geometry. For multiple-input multiple-output(MIMO) radar a solution utilizing a linear relationship in range-Doppler in order to classify the multipath ghosts, has been investigated previously. The focus was mainly on simulated data. Additionally, reconstruction of the reflecting surfaces, using the multipath propagation, has been done on solely simulated data. In this thesis we propose a similar solution, utilizing the linear relationship, adapted to handle real data. Particularly, reconstructing the reflecting surfaces is in focus. The aim of this thesis is to evaluate the solution in a range of scenarios, assessing its capacity to correctly reconstruct reflective surfaces. It is shown to handle varying angled single walls with good results. Scenes with multiple walls has also been investigated, showing promising results. With the reflective surfaces in a scene mapped, this can for example be used to designate areas where only multipath signals will occur, making it possible for these to be filtered out and ignored. The thesis has focused on human targets, but the solution has potential to be applicable in automotive scenarios as well, where the same linear relationship should be present.