<p>The characteristics of the point microcracks on the aeronautic ceramic bearing roller exhibit fuzzy and closed edges, with significant and concentrated depth. A coupling 3D morphological reconstruction method based on multi-restricted feature matching and semi-global stereo matching is established. To address the challenge of accurately correlating fuzzy feature points within closed regions, a dual-constraint matching strategy is proposed that combines inverse and guided matching. This approach effectively leverages the characteristics of fuzzy and closed edges. By analyzing the longitudinal and deep concentration of point microcracks, a multipath-aggregated semi-global matching model is developed that synthesizes high-resolution depth maps and generates dense point clouds.The experimental results show that the average reprojection error of the reconstructed point cloud is 0.01, the difference between the 3D model and the point microcrack extraction feature points fluctuates around 0.5, and the robust estimation validates 98% of the number of point clouds. The longitudinal and deep undulation characteristics of point microcracks are presented in the texture grid, and a 3D morphological reconstruction of point microcracks on the aeronautic ceramic bearing roller is successfully achieved. This method theoretically not only provides new ideas for the nondestructive characterization of complex geometric defects but also significantly improves the reliability of bearing fatigue life prediction. In addition, high-precision 3D characterization is particularly useful in potential application areas, such as the optimization of precision ceramic manufacturing processes. It provides key technical support for industrial equipment reliability analysis and life management.</p>