Abstract: The high-performance metal components manufactured by laser additive manufacturing have complex structures and surface conditions such as uneven roughness, oxide layer, and unfused particles, which cause scattering of detection signals, reduce the signal-to-noise ratio of crack reflection signals, mask small cracks,and affect the accuracy of detection results. Therefore, a GER-YOLO detection method for crack defects in high-performance metal components manufactured by laser additive manufacturing is proposed. By using digital sampling technology, combined with the defect development trends of different metal materials and the surface roughness of components, a discretization model of the surface roughness of metal components is established. Using the established model, the signal emission state is simulated to obtain the scattering law of detection signals, and the GER-YOLO network prediction head structure is constructed to extract the defect signals characteristics of metal components. X-rays are used to obtain detection signals, and the recognition effect of defect images is improved through contrast enhancement technology. The GER-YOLO detection model is optimized by introducing the coverage ratio threshold and Distribution Focal Loss function. The obtained model is used to analyze defect images and annotate crack defect characteristics. Finally, based on the annotated characteristics, an edge detection algorithm is used to obtain small defect image blocks, binarization and Canny operators are used to obtain the defect size, and the pulse reflection method is combined to determine the defect position, realizing the crack defect detection of additive manufacturing metal components. Experimental results conducted using the aforementioned design method demonstrated that it achieves a detection accuracy of over 70% for various crack defects, enabling precise identification of cracks in additive manufacturing metal components.