Report
3 search hits
-
Multi-View Reconstruction of Unknown Objects in the Presence of Known Occlusions
(2009)
-
Stefan Kuhn
Dominik Henrich
- We present a general method for reconstructing unknown objects (e.g. humans) within a known environment (e.g. tables, racks, robots) which usually has occlusions. These occlusions have to be considered since parts of the unknown objects might be hidden in some or even all camera views. Besides grayscale and color cameras also depth sensors are considered. In order to avoid cluttered reconstructions, plausibility checks are used to eliminate reconstruction artifacts which actually do not contain any unknown object. One application is a supervision/surveillance system for safe human/robot-coexistence and –cooperation. Experiments for a voxel-based implementation are given.
-
Multi-View Reconstruction in-between Known Environments
(2010)
-
Stefan Kuhn
Dominik Henrich
- We present a novel multi-view 3D reconstruction algorithm which unifies the advantages of several recent reconstruction approaches. Based on a known environment causing occlusions and on the cameras pixel grid discretization, an irregular partitioning of the reconstruction space is chosen. Reconstruction artifacts are rejected by using plausibility checks based on additional information about the objects to be reconstructed. The binary occupancy decision is solely performed in reconstruction space instead of fusing back-projected silhouettes in image space. Hierarchical data structures are used to reconstruct the objects progressively focusing on boundary regions. Thus, the algorithm can be stopped at any time with a certain conservative level of detail. Most parts of the algorithm may be processed in parallel using GPU programming techniques. The main application domain is the surveillance of real environments like in human/robot coexistence and cooperation scenarios.
-
Parallel Low-Storage Runge-Kutta Solvers for ODE Systems with Limited Access Distance
(2010)
-
Matthias Korch
Thomas Rauber
- We consider the solution of initial value problems (IVPs) of large systems of ordinary differential equations (ODEs) for which memory space requirements determine the choice of the integration method. In particular, we discuss the space-efficient sequential and parallel implementation of embedded Runge-Kutta (RK) methods. We focus on the exploitation of a special structure of commonly appearing ODE systems, referred to as "limited access distance", to improve scalability and memory usage. Such systems may arise, for example, from the semi-discretization of partial differential equations (PDEs). The storage space required by classical RK methods is directly proportional to the dimension n of the ODE system and the number of stages s of the method. We propose an implementation strategy based on a pipelined processing of the stages of the RK method and show how the memory usage of this computation scheme can be reduced to less than three storage registers by an overlapping of vectors without compromising the choice of method coefficients or the potential for efficient stepsize control. We analyze and compare the scalability of different parallel implementation strategies in detailed runtime experiments on different parallel architectures.
