Saupe, Dietmar
0000-0001-6735-5103 Berufsbeschreibung
Nachname
Vorname
Name
Suchergebnisse Publikationen
Complex 3D Shape Recovery using Hybrid Geometric Shape Features in aHierarchical Shape Segmentation Approach
We present a novel and reliable approach for complex object acquisition and surface registration using hybrid geometric shape features in a hierarchical 3D shape approximation and segmentation approach. First, instead of relying on one type of scanned data, we propose to use hybrid data provided that it can support both global and local geometric shape features. The scanned low-resolution global data supplies the global shape prior for registering the high-resolution local surface patches. Local surfaces can thus be optimally registered requiring less overlap and reducing uncertainty. Second, we cannot directly register huge volumes of data simultaneously due to the memory bottlenecks. We segment the global low-resolution model into several meaningful sub-shapes extending a hierarchical algorithm. The local surfaces can be registered on the sub-shapes respectively and all sub-shapes can be merged and rendered after registration. To verify the reliability of the approach, various 3D models have been acquired. The experiments show compelling results by reconstructing very detailed models of complex objects. The approach can be applied to practical 3D modeling applications.
Isometry-invariant matching of point set surfaces
Shape deformations preserving the intrinsic properties of a surface are called isometries. An isometry deforms a surface without tearing or stretching it, and preserves geodesic distances. We present a technique for matching point set surfaces, which is invariant with respect to isometries. A set of reference points, evenly distributed on the point set surface, is sampled by farthest point sampling. The geodesic distance between reference points is normalized and stored in a geodesic distance matrix. Each row of the matrix yields a histogram of its elements. The set of histograms of the rows of a distance matrix is taken as a descriptor of the shape of the surface. The dissimilarity between two point set surfaces is computed by matching the corresponding sets of histograms with bipartite graph matching. This is an effective method for classifying and recognizing objects deformed with isometric transformations, e.g., non-rigid and articulated objects in different postures.
On thinning methods for data assimilation of satellite observations
Thinning of observational data sets is an essential task in assimilation of satellite data for numerical weather forecast. In this work we modify and improve the scheme of so-called estimation error analysis (EEA). EEA is an adaptive thinning method that iteratively removes observations from a given data set, guided by a special approximation error measure evaluated at all original observation points. We propose EEA variants that differ in methodological and performance aspects, such as the Grid-EEA method, where errors are evaluated on a regular grid on the globe. Moreover, in the Top-Down EEA, we propose to construct the thinnings by an iterative point insertion strategy, which leads to an improved performance since the number of insertion steps is typically much smaller than the number of corresponding removal operations in EEA. We also provide an efficient implementation of the proposed methods yielding a significant acceleration of the standard EEA approach.
Analysis and visualization of space-time Variant parameters in endurance sports training
We develop methods for data acquisition, analysis, and visualization of performance parameters in endurance sports with emphasis on competitive cycling. The goal of our work is to provide methods to improve the performance of athletes in training and in competition. Measurements from a palette of devices, including common bike computers, GPS-recorders, power meters, and spiroergometric devices are combined requiring data fusion and synchronization. In addition, we consider integration of maps with elevation information and areal photographs together with recorded video footage of training courses. We also apply biofeedback methods that require complex data processing. Appropriate ways to jointly present such diverse data need to be developed for their analysis and visualization. A particular challenge is the information visualization of complex multi-dimensional signals of traveling sensors. For the simultaneous visualization of large information quantities we use the Powerwall at the University of Konstanz, i.e., a large high resolution display (dimensions 5.20 m x 2.15 m, pixel resolution 4640 x 1920) which offers the display of high resolution terrain data and maps together with static and dynamic parameter sequences from measured training rides or a running biofeedback simulation. The following topics are addressed among others: (1)adaptation of training course profile for laboratory biofeedback simulation, (2) learning of tactic approaches for a given training course profile, (3) evaluation of efficiency of different feedback parameters. The paper surveys some of the initial approaches, in cooperation with the Heart Institute Lahr/Baden, Germany.
A simulator for race-bike training on real tracks
We develop methods for data acquisition, analysis, modeling and visualization of performance parameters in endurance sports with emphasis on competitive cycling. For this purpose, we designed a simulator to facilitate the measurement of training parameters in a laboratory environment, to familiarize cyclists with unknown tracks, and to develop models for training control and performance prediction. The simulator is based on a Cyclus2 ergometer (RBM Elektronik-Automation GmbH, Leipzig, Germany), which provides a realistic cycling experience since one can mount arbitrary bikes and its elastic suspension even allows for a sway pedal stroke. The eddy current brake guarantees non-slipping transmission of a braking resistance up to 3000 W. Operating the Cyclus2 in the gradient mode, we impose arbitrary slopes by our own platform independent PC-based control software at a sampling rate of 2 Hz. The height profiles for various tracks were recorded using a commercial GPS device. The Cyclus2 has two major constraints with respect to simulating real tracks: We must focus on tracks without downhill accelerations since it has no engine and the eddy current brake requires a minimum rotation velocity of the flywheel to accurately generate the brake force. Therefore, we fixed the derailleurs to a heavy gear and mounted four electronic buttons to the handlebar which act like shift levers of virtual gears. Our software incorporates the virtual gear into the gradient so that the cyclist feels a correct resistance while the flywheel exceeds the minimum rotation velocity in all realistic uphill scenarios. Moreover, we can simulate arbitrary gears easily and record them over time. As the physical flywheel rotation is faster than in the simulation, our software must correct the related performance data. The simulation includes a video playback that is synchronized with the cyclist's current position on the track. In addition, time, distance, speed, cadence, heart rate, power and gears are monitored, a 2D-projection of the course gives feedback on the progress and a gradient profile indicates the slope in the surrounding of the current position. Comparative outdoor tests with an SRM power meter (Schoberer Rad Messtechnik, Welldorf, Germany) show that the simulator gives reasonable estimates for different pacing strategies (constant power/speed/heart rate). In future, we strive to integrate a more precise mechanical model (Martin et al., 1998), extend the palette of physiological measurements (oxygen consumption, ECG, lactate etc) and implement models for these parameters. The whole system shall indicate the optimum pacing strategy as Gordon derived in 2005 for simple models and synthetic data. Using sophisticated biofeedback visualization, cyclists shall be able to optimally prepare themselves even for unfamiliar tracks on our simulator.
Acquisition of Performance Parameters in Race-Bike Training
We develop methods for data acquisition, analysis, and visualization of performance parameters in endurance sports with emphasis on competitive cycling. For this purpose we created a bicycle simulator based on a Cyclus 2 ergometer and our own PC-based control software. The main components of the simulation are: - a computer controlled pedal resistance according to the height profile of a cycling track - the recording and visualization of training data measurements (speed, cadence, power, heart rate, heightprofile etc.) - and a video display of the cycling track that shows the current position. Our goal is to familiarize cyclists with unknown tracks and optimize their performance by means of training control and performance prediction based on physiological models
Stochastic optimization of multiple texture registration using mutual information
We consider the problem of simultaneously registering several images to a 3D model. We propose a global approach based on mutual information that extends previous methods to incorporate the color, and does not require segmentation or feature extraction. We give a stochastic model for joint optimization of multiple image-to-model alignment and we propose a heuristic to solve it. Experiments with synthetic models showed that our algorithm is robust to varying illumination and surface characteristics. Experiments with real data showed that we can achieve very good accuracy even for an object with highly specular surface, in moderate lighting conditions.
Efficient 3D Shape Acquisition and Registration Using Hybrid Scanning Data
We consider efficient 3D shape acquisition and surface registration using dissimilar laser range scanners. Most previous solutions do not use global shape information for optimal local surface registration. In this paper, we exploit the fundamental 3D scanning trade-off between the coverage of the global shape structure and numerous local surface patches to construct a hybrid laser scanning system provided that it can acquire both global and local shape information. The scanned low-resolution global shape data supplies the global shape structural prior for registering the high-resolution local 3D surface patches. Local surface patches can thus be optimally registered requiring less overlapping and thus reducing redundancy. To verify the feasibility of this system, we have implemented a prototype based on two laser range scanners, a hand-held one for the coarse global low-resolution model and a second stationary high-resolution line scanning system. This prototype system was evaluated for various real 3D models. Based on geometric data alone without using texture information, the results show that the proposed hybrid 3D scanning approach outperforms previous approaches in the presence of noise and outliers. The approach can be further applied to other practical 3D shape applications.
Foundations of 3D Digital Libraries : current approaches and urgent research challenges
3D documents are an indispensable data type in many important application domains such as Computer Aided Design, Simulation and Visualization, and Cultural Heritage, to name a few. The 3D document type can represent arbitrarily complex information by composing geometrical, topological, structural, or material properties, among others. It often is integrated with meta data and annotation by the various application systems that produce, process, or consume 3D documents. We argue that due to the inherent complexity of the 3D data type in conjunction with and imminent pervasive usage and explosion of available content, there is pressing need to address key problems of the 3D data type. These problems need to be tackled before the 3D data type can be fully supported by Digital Library technology in the sense of a generalized document, unlocking its full potential. If the problems are addressed appropriately, the expected benefits are manifold and may lead to radically improved production, processing, and consumption of 3D content.