Adaptive Bilateral Smoothing For a Point-Sampled Blade Surface

Abstract

Optical inspection of blades is important in computer vision and manufacturing automation. One problem commonly encountered is that the scanned point cloud may be polluted by noise, mainly from the scanning equipment. How to smooth the blade surface while preserving the thin-walled feature of leading/trailing edges is a challenging task. In this paper, we propose an adaptive bilateral method for the smoothing of a point-sampled surface. This paper is motivated by a bilateral filtering technique of a two-dimensional image. The basis of the method is the application of information entropy to distinguish density difference of point cloud. By minimizing defined smoothing density entropy and preserving density entropy, the optimal surface-smoothing factor and feature-preserving factor are calculated at each vertex. Applying the obtained factors, the objective of smoothing surface while preserving thin-walled feature is achieved, which overcomes the feature shrinking/shriveling problem. Our method does not require surface triangulation or curvature computation, and is very suitable for the object of a point-sampled surface with high-curvature feature or sharp features. Its robustness and efficiency are confirmed by experiments.