Automatic repair of damaged metallic components using Wire-Arc Additive Manufacturing (WAAM) is a crucial instrument for sustainable manufacturing. Optical 3D geometry acquisition systems and reverse engineering algorithms are major cornerstones of this proposed WAAM process chain, however, challenges remain regarding point cloud registration for the resulting large data sets. This paper presents point cloud stitching and matching approaches using descriptors, thus improving the feasibility regarding computational time. The proposed algorithms are tested and validated in a laboratory setup consisting of a 6-axis industrial welding robot including an integrated 3D geometry acquisition system based on structured-light projection.