Alberto Jaspe Villanueva

Automatic room detection and reconstruction in cluttered indoor environments with complex room layouts

We present a robust approach for reconstructing the main architectural structure of complex indoor environments given a set of cluttered 3D input range scans. Our method uses an efficient occlusion-aware process to extract planar patches as candidate walls, separating them from clutter and coping with missing data, and automatically extracts the individual rooms that compose the environment by applying a diffusion process on the space partitioning induced by the candidate walls. This diffusion process, which has a natural interpretation in terms of heat propagation, makes our method robust to artifacts and other imperfections that occur in typical scanned data of interiors. For each room, our algorithm reconstructs an accurate polyhedral model by applying methods from robust statistics. We demonstrate the validity of our approach by evaluating it on both synthetic models and real-world 3D scans of indoor environments. Graphical abstractDisplay Omitted HighlightsWe reconstruct an architectural model from a laser scanned indoor environment.Our algorithm can handle complex and highly concave room arrangements.It automatically detects all rooms without knowing the number of rooms in advance.Our pipeline can cope with occlusions and clutter using a robust heat diffusion process.An evaluation on artificial and real world data shows the accuracy of the method.

IsoCam: Interactive Visual Exploration of Massive Cultural Heritage Models on Large Projection Setups

We introduce a novel user interface and system for exploring extremely detailed 3D models in a museum setting. Three-dimensional models and associated information are presented on a large projection surface controlled by a touch-enabled surface placed at a suitable distance in front of it. Our indirect user interface, dubbed IsoCam, combines an object-aware interactive camera controller with an interactive point-of-interest selector and is implemented within a scalable implementation based on multiresolution structures shared between the rendering and user interaction subsystems. The collision-free camera controller automatically supports the smooth transition from orbiting to proximal navigation, by exploiting a distance-field representation of the 3D object. The point-of-interest selector exploits a specialized view similarity computation to propose a few nearby easily reachable interesting 3D views from a large database, move the camera to the user-selected point of interest, and provide extra information through overlaid annotations of the target view. The capabilities of our approach have been demonstrated in a public event attended by thousands of people, which were offered the possibility to explore submillimetric reconstructions of 38 stone statues of the Mont’e Prama Nuragic complex, depicting larger-than-life human figures, and small models of prehistoric Nuraghe (cone-shaped stone towers). A follow-up of this work, using 2.5m-high projection screens, is now included in permanent exhibitions at two Archeological Museums. Results of a thorough user evaluation, involving quantitative and subjective measurements, are discussed.

Robust Reconstruction of Interior Building Structures with Multiple Rooms under Clutter and Occlusions

We present a robust approach for reconstructing the architectural structure of complex indoor environments given a set of cluttered input scans. Our method first uses an efficient occlusion-aware process to extract planar patches as candidate walls, separating them from clutter and coping with missing data. Using a diffusion process to further increase its robustness, our algorithm is able to reconstruct a clean architectural model from the candidate walls. To our knowledge, this is the first indoor reconstruction method which goes beyond a binary classification and automatically recognizes different rooms as separate components. We demonstrate the validity of our approach by testing it on both synthetic models and real-world 3D scans of indoor environments.

Mont’e Scan: Effective Shape and Color Digitization of Cluttered 3D Artworks

We propose an approach for improving the digitization of shape and color of 3D artworks in a cluttered environment using 3D laser scanning and flash photography. To separate clutter from acquired material, semiautomated methods are employed to generate masks used to segment the range maps and the color photographs. This approach allows the removal of unwanted 3D and color data prior to the integration of acquired data in a 3D model. Sharp shadows generated by flash acquisition are easily handled by this masking process, and color deviations introduced by the flash light are corrected at the color blending step by taking into account the geometry of the object. The approach has been evaluated in a large-scale acquisition campaign of the Mont’e Prama complex. This site contains an extraordinary collection of stone fragments from the Nuragic era, which depict small models of prehistoric nuraghe (cone-shaped stone towers), as well as larger-than-life archers, warriors, and boxers. The acquisition campaign has covered 37 statues mounted on metallic supports. Color and shape were acquired at a resolution of 0.25mm, which resulted in more than 6,200 range maps (about 1.3G valid samples) and 3,817 photographs.

SSVDAGs: symmetry-aware sparse voxel DAGs

Voxelized representations of complex 3D scenes are widely used nowadays to accelerate visibility queries in many GPU rendering techniques. Since GPU memory is limited, it is important that these data structures can be kept within a strict memory budget. Recently, directed acyclic graphs (DAGs) have been successfully introduced to compress sparse voxel octrees (SVOs), but they are limited to sharing identical regions of space. In this paper, we show that a more efficient lossless compression of geometry can be achieved, while keeping the same visibility-query performance, by merging subtrees that are identical through a similarity transform, and by exploiting the skewed distribution of references to shared nodes to store child pointers using a variabile bit-rate encoding. We also describe how, by selecting plane reflections along the main grid directions as symmetry transforms, we can construct highly compressed GPU-friendly structures using a fully out-of-core method. Our results demonstrate that state-of-the-art compression and real-time tracing performance can be achieved on high-resolution voxelized representations of real-world scenes of very different characteristics, including large CAD models, 3D scans, and typical gaming models, leading, for instance, to real-time GPU in-core visualization with shading and shadows of the full Boeing 777 at sub-millimetric precision.

Reconstructing Complex Indoor Environments with Arbitrary Wall Orientations

Reconstructing the architectural shape of interiors is a problem that is gaining increasing attention in the field of computer graphics. Some solutions have been proposed in recent years, but cluttered environments with multiple rooms and non-vertical walls still represent a challenge for state-of-the-art methods. We propose an occlusionsaware pipeline that extends current solutions to work with complex environments with arbitrary wall orientations.

SOAR: Stochastic Optimization for Affine global point set Registration

We introduce a stochastic algorithm for pairwise affine registration of partially overlapping 3D point clouds with unknown point correspondences. The algorithm recovers the globally optimal scale, rotation, and translation alignment parameters and is applicable in a variety of difficult settings, including very sparse, noisy, and outlierridden datasets that do not permit the computation of local descriptors. The technique is based on a stochastic approach for the global optimization of an alignment error function robust to noise and resistant to outliers. At each optimization step, it alternates between stochastically visiting a generalized BSP-tree representation of the current solution landscape to select a promising transformation, finding point-to-point correspondences using a GPU-accelerated technique, and incorporating new error values in the BSP tree. In contrast to previous work, instead of simply constructing the tree by guided random sampling, we exploit the problem structure through a low-cost local minimization process based on analytically solving absolute orientation problems using the current correspondences. We demonstrate the quality and performance of our method on a variety of large point sets with different scales, resolutions, and noise characteristics.

PEEP: perceptually enhanced exploration of pictures

We present a novel simple technique for rapidly creating and presenting interactive immersive 3D exploration experiences of 2D pictures and images of natural and artificial landscapes. Various application domains, ranging from virtual exploration of works of art to street navigation systems, can benefit from the approach. The method, dubbed PEEP, is motivated by the perceptual characteristics of the human visual system in interpreting perspective cues and detecting relative angles between lines. It applies to the common perspective images with zero or one vanishing points, and does not require the extraction of a precise geometric description of the scene. Taking as input a single image without other information, an automatic analysis technique fits a simple but perceptually consistent parametric 3D representation of the viewed space, which is used to drive an indirect constrained exploration method capable to provide the illusion of 3D exploration with realistic monocular (perspective and motion parallax) and binocular (stereo) depth cues. The effectiveness of the method is demonstrated on a variety of casual pictures and exploration configurations, including mobile devices.