An Estimation Algorithm of the Axis of Rotation from 3D Cloud Data Based on Line Element

Because of the advantages on accuracy and expeditiousness in recent CAD modeling and reverse engineering, feature-based surface reconstruction becomes a hot topic in reverse engineering. Especially, 3D shape recognition and feature extraction is the difficulty and key point of the feature-based reverse engineering. To improve the performance and robustness, a new method for the estimation of the axis of rotation based on line element geometry, line complex and kinematic equation is presented. The method first projects the data points of a 3D volume to a line element image-space. A linear complex of the line elements is built and then the axis is approximated by fitting a kinematic equation on it. It is more efficient and faster than the previous methods because it doesnt need to estimate accurate surface normal vector of the cloud data. A new discrete region growing technique, called K-Local-RANSAC, is adopted to exclude the influence of the outer points and ensure the robustness. Line element geometry is employed to effectively perform the presentation of complex objects according to surface type. The K-Local-RANSAC algorithm uses linear complex to perform the classification. The discrete region growing mode improves the efficiency obviously. However, the effectivity on general rotational surface including torus and noise and fragment cloud points has been proved by experiments.