In this work, two approaches for computing the near-field distribution around wind turbines are proposed, including: (1) Huygens Principle and (2) the parabolic equation technique. In order to simplify the problem, the cylinder model is utilized to represent the wind turbines and transform the problem into a two-dimensional case. To make Huygens Principle computationally tractable, several approximations are made based on the problem geometry especially modelling the cylinder as a plate. The expression of the electromagnetic field radiated by the equivalent magnetic current can be analytically solved by the error function. To verify the results, FEKO is utilized to simulate the scattering of infinitely long cylinders using periodic boundary condition (PBC). In order to solve the problem of multiple cylinders, a modified method is derived. For more accurate results, the parabolic equation (PE) technique is utilized to solve this problem, which is usually utilized to solve wave propagation problems. In this case, wide-angle approximation is used to solve the parabolic equation, which can obtain accurate results in a region of up to 45 degrees. Although these two approaches are not full-wave simulation, the calculation time is significantly reduced and the error is acceptable. To further verify the computed results by the parabolic equation technique, two commercial transceivers from Time Domain Corporation are used to measure the field distribution behind a finite-length metal pole. The frequency-domain results are obtained from the measured time-domain results using the fast Fourier transform. It is shown that the computed results by the parabolic equation technique agree well with the measurement results. / text

Le travail présenté dans ce mémoire a pour objectif le développement de nouvelles approches permettant d'introduire de manière robuste des marques dans une vidéo. Il s'agira tout d'abord de sélectionner les zones des images les plus appropriées à l'introduction du tatouage. Pour cela, plusieurs critères devront être pris en compte. Tout d'abord, lorsqu'une marque est introduite dans une zone de l'image, il est nécessaire d'introduire la même marque au même endroit sur toutes les images de la séquence. En effet, dans le cas contraire, un simple filtrage temporel dans le sens du mouvement permet de détruire la marque. Il sera également important de détecter les zones moins importantes visuellement dans lesquelles des modifications pourraient être introduites plus facilement sans qu'elles soient visibles. Il s'agit notamment des zones visibles dans peu d'images différentes, ou des zones ou l'œil humain est peu susceptible de détecter des différences, telles que les zones d'ombres en mouvement. Pour parvenir à une détection efficace de telles régions, une phase d'analyse permettant notamment la création d'images mosaïques (ou sprite dans le contexte MPEG-4) et la détection des ombres en mouvement devra être réalisée. / The work presented in this thesis aims to develop new approaches to introduce, in a robust way, a marks in a video. The areas of the images which are most appropriate to the introduction of the tattoo should be selected. For that, several criteria must be taken into account. Firstly, when a mark is introduced into an area of the image, it is necessary to introduce the same mark at the same location on all images in the sequence. It is also important to detect an invisible areas where changes could be introduced easily without being seen. These are areas where the human eye is less sensitive to changes, such as shadows region of moving objects. To achieve the e#ective detection of such regions, a phase of analysis including the creation of mosaic images and the detection of shadows of moving objects will be realized.