Based on the non-contact nature and penetration capability of continuous-wave terahertz imaging， this study proposed a nondestructive detection method for metallic foreign objects in energetic materials.Terahertz imaging experiments were conducted on four types of metallic foreign objects covered by energetic-material layers with thicknesses of 0.5-2.0 cm， and a corresponding terahertz image dataset was established. Based on this dataset， a lightweight recognition model based on YOLO26-Partial was developed for metallic foreign object detection. In addition， an enhanced generative adversarial network， termed EDGAN， was proposed for terahertz image super-resolution reconstruction by integrating an attention mechanism with a deep residual structure. Experimental results demonstrate that the proposed recognition algorithm achieves an average recognition accuracy of 99% for the four types of metallic foreign objects while reducing model complexity. Compared with ESRGAN， the proposed reconstruction algorithm achieves average improvements of 5.70% and 1.36% in peak signal-to-noise ratio （PSNR） and structural similarity index （SSIM）， respectively. The proposed method enables both accurate recognition and high-quality super-resolution reconstruction of metallic foreign objects in energetic materials， providing a feasible technical approach for nondestructive inspection in related applications.