Reconstruction of Mixed Boundary Objects and Classification Using Deep Learning and Linear Sampling Method

Abstract

The linear sampling method is a simple and reliable linear inversion technique for determining the morphological features of unknown objects under investigation. Nevertheless, there are many challenges that this method depends on the frequency of operation and it is unable to produce satisfactory results for objects with complex shapes. This paper proposes a hybrid model, which combines conventional linear sampling method and deep learning for the reconstruction of mixed boundary objects. In this approach, the initial approximation of mixed boundary objects derived from linear sampling method serves as the training data for the U-Net based convolutional neural network. The network then learns to correlate this approximation with the corresponding ground truth profiles. Along with the reconstruction of mixed boundary objects, they are also classified as dielectric or conductor, and count of each object type are measured. Furthermore, the low-frequency and high-frequency characteristics of the linear sampling method are analyzed, and its limitations are overcome by combining it with a deep learning approach. The effectiveness of the proposed model is validated using several examples of synthetic and experimental data. The results demonstrate that the proposed method outperforms the conventional Linear sampling method in terms of accuracy.