Carbon fiber reinforced polymer cables have been used due to their excellent properties, making surface damage detection an essential task. This paper investigates damage detection in carbon fiber cables using electromagnetic tomography, focusing on the influence of defect orientation on the response of the inverse problem reconstruction signal. A novel EMT image reconstruction algorithm is proposed, integrating multi-frequency excitation with Elastic Net regularization. The results demonstrate that for surface defects, the induced of defect orientation on the induced voltage response is primarily governed by the anisotropic electrical conductivity, whereas near-surface defects exhibit a more complex directional dependence. The proposed algorithm successfully reconstructs images of two typical defects—debonding and broken strands—achieving an average MSE reduction of 84.8% and an average SSIM improvement of 28.3% compared to the conventional LBP algorithm. These results underscore the algorithm’s potential for defect imaging in carbon fiber cables using electromagnetic tomography.