Abstract <p>The inspection and characterization of surface cracks are important issues in the structural health monitoring (SHM) of structural components. Electromagnetic acoustic transducer (EMAT) is a nondestructive testing technology widely used in health monitoring. However, it is difficult to evaluate the depth of defects accurately with consideration of the ambiguous relationship between the cracks and feature values. Therefore, an accurate evaluation method for the depth of surface cracks needs to be established. In this paper, a fusion characterization method based on fuzzy relationship matrices is proposed. Firstly, the different depth estimated values of the surface crack are obtained based on the relationship modals between the crack depth and multiple different feature values. Secondly, a difference matrix is established that can describe the difference in accuracy between different depth estimated values, and a relationship matrix is constructed based on the triangular fuzzy number method for evaluating the quantization accuracy. Thirdly, the weights of different estimated values in the fusion quantification were determined, and the final defect depth was obtained. Finally, we applied our approach to experimental data and its maximum quantization relative error was 5.9%, which demonstrates its capability to characterize surface cracks compared with other methods.</p>