This paper presents a novel integrated framework for high-precision aero-engine blade damage detection and quantification through advanced 3D reconstruction and point cloud registration techniques. Traditional inspection methods suffer from limitations, including manual dependency, 2D constraints, and environmental sensitivity. To address these challenges, we propose a comprehensive approach that combines COLMAP structure-from-motion with CasMVSNet for high-fidelity 3D reconstruction, implements a robust multi-stage denoising strategy integrating K-means clustering with score-based refinement, and introduces an enhanced Scale-Invariant Feature Transform (SIFT) and 3D Shape Context (3DSC) fusion descriptor for reliable feature matching. The framework employs the Generalized Iterative Closest Point (GICP) algorithm for precise point cloud registration between damaged and intact blade models. The illustrated results have demonstrated that the proposed method achieves robust geometric comparison capabilities under challenging conditions, including noise and geometric deformation, providing a reliable solution for intelligent aero-engine blade damage detection.

Aero-engine blade,3D reconstruction,point cloud denoising,point cloud registration