The use of computer graphics techniques in cultural heritage (CH) has led to impressive improvements in technologies related to digital acquisition and rendering of 3D CH data. Digitized artefacts are becoming widely available for access and reuse, thus increasing the need of tools able to support comparative shape analysis. As 3D artefacts are often worn, eroded and broken, these tools cannot take advantage of existing methods based on exact matching but they rather require new approaches able to identify partial features in portions of models thus leading to a double partiality of the matching problem, in terms of both features and models In this context, we propose a method based on a novel generalization of the Hough transform technique able to identify and localize semantic features like anatomical features, ornaments, or decorations on digital artefacts or fragments, even if the features are partially damaged or incomplete. The major advantages of using a method based on the Hough transform technique are the relative robustness to noise and the recognition power also in the case of partial features. Our experiments on digital models of real artefacts are encouraging and show the potential of the method, which can work on both 3D meshes and point clouds.