Template-Guided 3D Fragment Reassembly Using GDS

Abstract

Computer-aided fragment reassembly becomes more and more significant in recent years. The state of the art methods mainly utilize the fracture surface of the fragment. However, some fracture surfaces are often eroded and the features are not discriminative enough for matching. In this paper, we proposed a template-guided 3D fragment reassembly algorithm using Geodesic Disk Spectrum (GDS), which conducts matching between the intact surface of the fragment and the template. A two-step procedure is proposed for the first time with GDS-based matching and ICP-based registration for the reassembly task. The largest enclosed geodesic disk of the fragment is extracted and the matching to the template is found by GDS. In order to reduce the computational complexity, a k-layer Normal Distribution Descriptor (NDD) is also proposed. Transformation of the matched geodesic disks is obtained using the Iterative Closest Points (ICP) algorithm, and the registration between the fragment and the template is achieved. Our algorithm has been tested on various fragments and accurate results are obtained. A higher precision is achieved by comparing with existing algorithms, which proves the efficiency.