Bottom-Up Clustering and Merging Strategy for Irregular Curvature Aero-Blade Surface Extraction

Abstract

Surface extraction plays a significant role in blade reconstruction and measurement, improving manufacturing quality and reducing costs. Boundary extraction and region growth are the most common solutions. However, these methods treat points as complete objects and extract surfaces aggressively. Since the model and geometric constraints of the surfaces are usually blurring, these methods will lead to low accuracy results and dramatically impact the effectiveness of existing methods in practice. This paper presents a novel bottom-up irregular curvature surface extracting method called SuCMIS segmenting blade surface points from an unstructured scanning point cloud. SuCMIS applies adaptive multiple clustering methods based on the local flatness and normal deviation minimum, which improves the robustness to noise and unpredicted systematic errors. Then the curvature-based cluster merging process runs to reconstruct the blade surface. Differ to existing model-based method, SuCMIS converts surface extraction into a bottom-up process with multiply surfaces clustering and merging. It does not rely on evaluating the accuracy of the model and geometric constraints, which overcomes the natural defects of existing methods significantly. Experiments demonstrate SuCMIS’s efficiency and robustness. Compared to baseline methods, SuCMIS improves extracting accuracy with acceptable execution time, especially for complicated and large point clouds, such as blade surfaces.