In many industrial design applications, it is necessary to track the positions and orientations of a static object in 3D space. This paper presents a method to quickly and accurately locate the placements of an object based on 3D scanning techniques. By registering geometric data from different scans, the affine transformation from the base position to an arbitrary position within the scanner working envelope can be found. Depending on differen types of applications, the placements of the scanner, the object and a reference object are discussed. Based on a series of experiments, the robustness of the presented method is verified and the errors, both in translation and in rotation, are identified. The possible applications of the presented method in industrial design are investigated as well.@en

We present a method of partial matching two shapes aimed to support 3D scanning applications. To digitize the surface of a 3D object, a number of scan views are taken from different angles, collectively representing the entire surface of the object. The registration process assigns a transformation to each of the scan views such that the surface of the object can be reconstructed. The motivation of our work is to provide an initial geometric model to a designer who develops the shape further using Computer-Aided Design. We propose a method of pairwise shape matching of partial scans, without depending on any distinct geometric feature type. The degree of matching of two shapes, that is the congruence of their overlapping portions, has been defined by a matching index I, which should be maximized to attain the right relative placement of the two shapes. By using a simplification of one of the shapes and a sampling strategy in 6D configuration space we achieve matching in an amount of time that would be too high for practical interactive use. However, we offload the most computation intensive part of the matching process to the graphical processing unit (GPU) to achieve 50-80 times shorter times. The method does not require the detection of intrinsic or artificial features (or shape descriptors), nor optical markers nor position or orientation tracking information. To our best knowledge there are no previous reports about practically usable results of similar methods. We present the outcomes of tests with scanned point clouds and initial conclusions about the robustness and costs of the method. The limitations of the method are discussed and improvements of the search strategy and the GPU-based computation of the matching index are proposed.@en

We present a method of automatic alignment of two partially matching 3D shapes. The algorithm selects large fat tetrahedrons (LFT) composed of 4 vertices in one shape and exhaustively searches in the other shape for sets of 4 vertices being compatible with the tetrahedron. By selecting such salient tetrahedrons that are relatively wide and fat, although also being not too unlikely to be contained in the overlap region, the cost of search can be reduced. The method is relatively insensitive to noise and not depending on the existence of local shape features nor on feature correspondences. When implemented on a GPU in Cuda, two point sets of 40,000 each can be aligned within seconds. The method is intended to support interactive 3D scan registration applications.@en