Structure recovery by part assembly

Abstract

Figure 1: Given single-view scans by the Kinect system, containing highly noisy and incomplete 3D scans (upper left) and corresponding RGB images (lower left), our approach is able to faithfully recover their underlying structures (yellow) by assembling suitable parts (red) in the repository models (blue). AbstractThis paper presents a technique that allows quick conversion of acquired low-quality data from consumer-level scanning devices to high-quality 3D models with labeled semantic parts and meanwhile their assembly reasonably close to the underlying geometry. This is achieved by a novel structure recovery approach that is essentially local to global and bottom up, enabling the creation of new structures by assembling existing labeled parts with respect to the acquired data. We demonstrate that using only a small-scale shape repository, our part assembly approach is able to faithfully recover a variety of high-level structures from only a single-view scan of man-made objects acquired by the Kinect system, containing a highly noisy, incomplete 3D point cloud and a corresponding RGB image. Links:DL PDF WEB

Recovering Functional Mechanical Assemblies from Raw Scans

“…Our work also shares resemblance to the large body of methods on structure-based shape analysis and modeling [17], [18], [19], [20], [21], [22], [23], wherein high-level shape structures are identified with low-dimensional parts of different sizes, positions, and interactions. Gal et al [17] demonstrated that working with a set of 1D feature curves extracted from engineered objects, and preserving their intra-and inter-relations while deforming the shapes lead to an intuitive manipulation framework.…”

“…Gal et al [17] demonstrated that working with a set of 1D feature curves extracted from engineered objects, and preserving their intra-and inter-relations while deforming the shapes lead to an intuitive manipulation framework. Shen et al [19] recovered structures by combining existing labeled parts from different objects, considering both partial matching with the acquired data and the interactions between parts. Laga et al [21] exploited the context of structural relationships between shape parts and use them, in conjunction to their geometry, as cues for functionality recognition.…”

“…We adopt a similar strategy as in [19] to perform shape matching between the template and the part. The normals of their largest planes are firstly aligned.…”

Ergonomics-Inspired Reshaping and Exploration of Collections of Models

“…For an extensive discussion of history and development of ergonomics in the context of chairs we recommend Cranz's book [17], especially Chapters 3 and 5. As examples of linking geometric form with function, Shen et al [18] utilize object structure for effective object assembly, while the AttribIt [19] system links semantic tags to object parts and their coupled variations.…”

Learning Hierarchical Semantic Segmentations of LIDAR Data

“…[28] uses 3D CAD models as templates for a sliding window search. Shen et al [25] use a database of segmented models and assembly-based modeling algorithms to reconstruct novel 3D models. Nan et al [20] integrate a region growing segmentation algorithm with object recognition for cluttered indoor scene understanding.…”