Terahertz-based super-resolution imaging and quantitative evaluation method for defects in composite materials

Composite materials play a critical role in advanced engineering applications, and reliable non-destructive evaluation is essential for ensuring the integrity and safety of their service performance. To meet the demand for high-precision characterization of subsurface defects in glass fiber reinforced polymer (GFRP) laminates, this study proposes a terahertz (THz) image super-resolution method and a physics-guided quantitative evaluation framework. A Helmholtz-constrained dual-path convolutional neural network (CNN) is developed, in which the Helmholtz equation is embedded as a physical regularization term to enhance spatial fidelity and suppress false scattering artifacts. A biomimetic compound-eye-inspired feature extraction strategy is further introduced and integrated with multi-scale Hessian filtering and channel attention mechanisms, enabling the construction of a multidimensional damage-assessment matrix for correlation-based quantification of defect morphology and fiber-related structural features. In addition, Terahertz–Frequency-Modulated Continuous Wave (THz-FMCW) data processing and 3D imaging are incorporated to provide supplementary depth information. Experimental results demonstrate that the proposed method reconstructs defect geometry and electromagnetic scattering characteristics with improved fidelity under sparse sampling conditions. The framework also yields more consistent morphological descriptors and spatial metrics compared with conventional ultrasonic inspection. These findings show that the presented approach offers a physically grounded and high-precision pathway for advancing terahertz-based non-destructive evaluation of composite structures.