부산대학교 도서관

In microwave imaging (MI), using prior boundary information to enhance the accuracy of retrieved dielectric properties holds great promise. However, the precise registration of prior boundaries with the actual boundaries of the imaged object is crucial, as misalignment can result in significant inaccuracies in retrieved values. To address this issue, we present a novel approach that formulates the problem of aligning the prior and actual boundaries as a shape optimization problem. The proposed method uses the free-form deformation (FFD) technique to parameterize the shape optimization problem. Furthermore, we introduce a new method for efficient computation of the Jacobian matrix by analyzing the sensitivity of the scattered field with respect to the location of control points in the FFD. Through simulations of MI of the lower extremity of the human body, we demonstrate the effectiveness of our proposed approach. Our results indicate that the proposed approach reduces the average Hausdorff distance (AHD) between the prior and actual boundaries by approximately 80%, leading to a substantial enhancement in the accuracy of retrieved dielectric properties.