An official website of the United States government
We introduce a method for high-quality 3D reconstruction from multi-view images. Our method uses a new point-based representation, the regularized dipole sum, which generalizes the winding number to allow for interpolation of per-point attributes in point clouds with noisy or outlier points. Using regularized dipole sums, we represent implicit geometry and radiance fields as per-point attributes of a dense point cloud, which we initialize from structure from motion. We additionally derive Barnes-Hut fast summation schemes for accelerated forward and adjoint dipole sum queries. These queries facilitate the use of ray tracing to efficiently and differentiably render images with our point-based representations, and thus update their point attributes to optimize scene geometry and appearance. We evaluate our method in inverse rendering applications against state-of-the-art alternatives, based on ray tracing of neural representations or rasterization of Gaussian point-based representations. Our method significantly improves 3D reconstruction quality and robustness at equal runtimes, while also supporting more general rendering methods such as shadow rays for direct illumination.
more »
« less
Globally Consistent Normal Orientation for Point Clouds by Regularizing the Winding-Number Field
Estimating normals with globally consistent orientations for a raw point cloud has many downstream geometry processing applications. Despite tremendous efforts in the past decades, it remains challenging to deal with an unoriented point cloud with various imperfections, particularly in the presence of data sparsity coupled with nearby gaps or thin-walled structures. In this paper, we propose a smooth objective function to characterize the requirements of an acceptable winding-number field, which allows one to find the globally consistent normal orientations starting from a set of completely random normals. By taking the vertices of the Voronoi diagram of the point cloud as examination points, we consider the following three requirements: (1) the winding number is either 0 or 1, (2) the occurrences of 1 and the occurrences of 0 are balanced around the point cloud, and (3) the normals align with the outside Voronoi poles as much as possible. Extensive experimental results show that our method outperforms the existing approaches, especially in handling sparse and noisy point clouds, as well as shapes with complex geometry/topology.
more »
« less
- Award ID(s):
- 2007661
- PAR ID:
- 10558866
- Publisher / Repository:
- ACM
- Date Published:
- Journal Name:
- ACM Transactions on Graphics
- Volume:
- 42
- Issue:
- 4
- ISSN:
- 0730-0301
- Page Range / eLocation ID:
- 1 to 15
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
In the plane, thewinding numberis the number of times a curve wraps around a given point. Winding numbers are a basic component of geometric algorithms such as point-in-polygon tests, and their generalization to data with noise or topological errors has proven valuable for geometry processing tasks ranging from surface reconstruction to mesh booleans. However, standard definitions do not immediately apply on surfaces, where not all curves bound regions. We develop a meaningful generalization, starting with the well-known relationship between winding numbers and harmonic functions. By processing the derivatives of such functions, we can robustly filter out components of the input that do not bound any region. Ultimately, our algorithm yields (i) a closed, completed version of the input curves, (ii) integer labels for regions that are meaningfully bounded by these curves, and (iii) the complementary curves that do not bound any region. The main computational cost is solving a standard Poisson equation, or for surfaces with nontrivial topology, a sparse linear program. We also introduce special basis functions to represent singularities that naturally occur at endpoints of open curves.more » « less
-
null (Ed.)This paper develops an analytic framework to design both stress-controlled and displacement-controlled T -periodic loadings which make the quasistatic evolution of a one-dimensional network of elastoplastic springs converging to a unique periodic regime. The solution of such an evolution problem is a function t ↦( e ( t ), p ( t )), where e i ( t ) is the elastic elongation and p i ( t ) is the relaxed length of spring i , defined on [ t 0 , ∞ ) by the initial condition ( e ( t 0 ), p ( t 0 )). After we rigorously convert the problem into a Moreau sweeping process with a moving polyhedron C ( t ) in a vector space E of dimension d , it becomes natural to expect (based on a result by Krejci) that the elastic component t ↦ e ( t ) always converges to a T -periodic function as t → ∞ . The achievement of this paper is in spotting a class of loadings where the Krejci’s limit doesn’t depend on the initial condition ( e ( t 0 ), p ( t 0 )) and so all the trajectories approach the same T -periodic regime. The proposed class of sweeping processes is the one for which the normals of any d different facets of the moving polyhedron C ( t ) are linearly independent. We further link this geometric condition to mechanical properties of the given network of springs. We discover that the normals of any d different facets of the moving polyhedron C ( t ) are linearly independent, if the number of displacement-controlled loadings is two less the number of nodes of the given network of springs and when the magnitude of the stress-controlled loading is sufficiently large (but admissible). The result can be viewed as an analogue of the high-gain control method for elastoplastic systems. In continuum theory of plasticity, the respective result is known as Frederick-Armstrong theorem.more » « less
-
Context. LDN 1157 is one of several clouds that are situated in the cloud complex LDN 1147/1158. The cloud presents a coma-shaped morphology with a well-collimated bipolar outflow emanating from a Class 0 protostar, LDN 1157-mm, that resides deep inside the cloud. Aims. The main goals of this work are (a) mapping the intercloud magnetic field (ICMF) geometry of the region surrounding LDN 1157 to investigate its relationship with the cloud morphology, outflow direction, and core magnetic field (CMF) geometry inferred from the millimeter- and submillimeter polarization results from the literature, and (b) to investigate the kinematic structure of the cloud. Methods. We carried out optical ( R -band) polarization observations of the stars projected on the cloud to map the parsec-scale magnetic field geometry. We made spectroscopic observations of the entire cloud in the 12 CO, C 18 O, and N 2 H + ( J = 1–0) lines to investigate its kinematic structure. Results. We obtained a distance of 340 ± 3 pc to the LDN 1147/1158, complex based on the Gaia DR2 parallaxes and proper motion values of the three young stellar objects (YSOs) associated with the complex. A single filament of ~1.2 pc in length (traced by the Filfinder algorithm) and ~0.09 pc in width (estimated using the Radfil algorithm) is found to run throughout the coma-shaped cloud. Based on the relationships between the ICMF, CMF, filament orientations, outflow direction, and the hourglass morphology of the magnetic field, it is likely that the magnetic field played an important role in the star formation process in LDN 1157. LDN 1157-mm is embedded in one of the two high-density peaks detected using the Clumpfind algorithm. The two detected clumps lie on the filament and show a blue-red asymmetry in the 12 CO line. The C 18 O emission is well correlated with the filament and presents a coherent structure in velocity space. Combining the proper motions of the YSOs and the radial velocity of LDN 1147/1158 and an another complex, LDN 1172/1174, that is situated ~2° east of it, we found that the two complexes are moving collectively toward the Galactic plane. The filamentary morphology of the east-west segment of LDN 1157 may have formed as a result of mass lost by ablation through interaction of the moving cloud with the ambient interstellar medium.more » « less
-
To study the sensing mechanism of bat's biosonar system, we propose a fast simulation algorithm to generate natural-looking trees and forest---the primary living habitat of bats. We adopt 3D Lindenmayer system to create the fractal geometry of the trees, and add additional parameters, both globally and locally, to enable random variations of the tree structures. Random forest is then formed by placing simulated trees at random locations of a field according to a spatial point process. By employing a single algorithmic model with different numeric parameters, we can rapidly simulate 3D virtual environments with a wide variety of trees, producing detailed geometry of the foliage such as the leaf locations, sizes, and orientations. Written in C++ and visualized with openGL, our algorithm is fast to implement, easily parallable, and more adaptive to real-time visualization compared with existing alternative approaches. Our simulated environment can be used for general purposes such as studying new sensors or training remote sensing algorithms.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.1145/3592129
