Enhanced Contrast Born Iterative Cascaded Network for High-Contrast Inverse Scattering Imaging

• Published in: IEEE antennas and wireless propagation letters pp. 1 - 5
• Main Authors: Wang, Jingjing, Li, Zhe, Xu, Huaqiang, Hu, Nannan
• Format: Journal Article
• Language: English
• Published: IEEE 2025
• Subjects: Accuracy; Coordinate Attention Mechanism; Enhanced Contrast; Image edge detection; Imaging; Inverse problems; Inverse Scattering Imaging; Iterative methods; Magnetic field measurement; Optimization; Radar imaging; Scattering; Training
• DOI: 10.1109/LAWP.2025.3593269
• ISSN: 1536-1225

Electromagnetic inverse scattering imaging retrieves information on the spatial position and material properties of the target based on the measured scattering field data, which has been widely used in non-destructive evaluation and wireless communication. However, high-contrast scatterers with severe nonlinearity bring about multiple scattering effects, which inevitably limit the imaging accuracy in edges and internal change regions. To address the problem, this letter proposes a novel Enhanced Contrast Born Iterative Cascade Network (ECBIC-Net) for high contrast inverse scattering imaging. In ECBIC-Net, the contrast of scatterers is first constrained by adding additional physical constraint factors to alleviate the multiple scattering effect. Then, an improved U-Net cascade network structure based on coordinate attention mechanism is used to improve the location-awareness of the network. Finally, a new multi-iteration weighted loss function is proposed to help the network balancing the learning process. Experimental results show that the proposed method can achieve normalized mean square error of 0.015 and structure similarity index measure of 0.905 on the dataset, which is superior to other reported methods.