Abstract: This talk presents a method that uses the level sets of volumes to reconstruct the shapes of 3D objects from range data. The strategy is to formulate 3D reconstruction as a statistical problem: find that surface which is mostly likely, given the data and some prior knowlege about the application domain. The resulting optimization problem is solved by an incremental process of deformation. We represent a deformable surface as the level set of a discretely sampled scalar function of 3 dimensions, i.e. a volume. Such level-set models have been shown to mimic conventional deformable surface models by encoding surface movements as changes in the greyscale values of the volume. The result is a voxel-based modeling technology that offers several advantages over conventional parametric models, including flexible topology, no need for reparameterization, concise descriptions of differential structure, and a natural scale space for hierarchical representations. This work builds on previous work in both 3D reconstruction and level-set modeling. It presents a fundamental result in surface estimation from range data: an analytical characterization of the surface that maximizes the posterior probability. It also presents a novel computational technique for level-set modeling, called the sparse-field algorithm, which combines the advantages of a level-set approach with the computational efficiency and accuracy of a parametric representation. The sparse-field algorithm is more efficient than other approaches, and because it assigns the level set to a specific set of grid points, it positions the level-set model more accurately than the grid itself. These properties, computational efficiency and sub-cell accuracy, are essential when trying to reconstruct the shapes of 3D objects. Results are shown for the reconstruction objects from sets of noisy and overlapping range maps.
Abstract: The needs of our society are imposing ever increasing demands for efficient intelligent methodologies in control and automation, signal processing and communication networks. It is important to learn how to build systems that exhibit high degrees of autonomous behavior. To rise to the challenge, significant progress in the area of hybrid systems is needed. Hybrid systems arise from the interaction of discrete planning algorithms and continuous processes, and as such, they provide the basic framework and methodology for the analysis and synthesis of autonomous and intelligent systems.
Hybrid systems contain two distinct types of components, subsystems with continuous dynamics and subsystems with discrete event dynamics, that interact with each other. Such systems are important in a variety of contexts: Hybrid systems frequently arise from computer aided control of continuous processes in manufacturing, communication networks, autopilot design, computer synchronization, traffic control, and industrial process control, for example. Another important way in which hybrid systems arise is from the hierarchical organization of complex control systems. In these systems, a hierarchical organization helps manage complexity and higher levels in the hierarchy require less detailed models (discrete abstractions) of the functioning of the lower levels, necessitating the interaction of discrete and continuous components. Examples of such systems include flexible manufacturing and chemical process control systems, interconnected power systems, intelligent vehicle highway systems, air traffic management systems, computer communication networks.
The study of hybrid control systems is essential in designing sequential supervisory controllers for continuous systems, and it is central in designing intelligent control systems with a high degree of autonomy. The investigation of hybrid systems is creating a new and fascinating discipline bridging control engineering, mathematics and computer science.
In this talk, after an introduction to intelligent and hybrid systems, analysis and synthesis methodologies for hybrid control systems will be discussed. Discrete event systems will also be further discussed using Petri nets and a computationally efficient approach for the design of supervisory controllers, that uses the place invariants of the net will be presented.
Abstract: The symmetry set and its close relative the medial axis has been in use for some years as a way of constructing a skeleton for a 2-dimensional shape, a sort of "curved axis of symmetry". In the talk I shall describe some recent attempts by Guillermo Sapiro (Electrical Engineering at UMN, Twin Cities) and myself to find analogous constructions which are invariant under affine (linear) transformations in the plane. The talk will presuppose nothing beyond a knowledge of calculus and linear algebra (and not much of those!). This is work in progress and we don't know very much as yet but I shall talk about two constructions---there are other possible ones---which have some nice features.
Abstract: A heretofore unsolved challenge is the completely automatic and accurate estimation of road boundaries in aerial images when the roads may be partially or completely locally occluded and clutter may be prevalent. In this talk I introduce a roadfinder that is effective in meeting this challenge. The roadfinder begins with one or more seeds on each long road, and then accurately estimates the remaining boundaries, which can be found completely automatically by the algorithm. The algorithm is robust to missing boundary edges on one side of the road and on both sides of the road simultaneously. These arise from shadows and occlusion by trees, small structures, etc. It is also robust to clutter within the road caused by cars or trucks, and to clutter resulting from intersecting or close parallel roads. The algorithm is based on simple clutter and occlusion models and a combined Multihypothesis Generalized Kalman Filter (MGKF).
Abstract: Alpha microprocessors have maintained leadership performance since their introduction in 1992. Three generations of microprocessors have been designed by an experienced, highly skilled design team using a proven design methodology. These microprocessors achieve performance by focusing on high frequency design. To facilitate this type of design, the Alpha instruction set architecture (ISA) uses simple fixed length instructions. In addition, Digital's CMOS technologies include specific features to enable high frequency, low-skew clock distribution. Complex circuit styles are used throughout these designs to meet aggressive cycle time goals. Close interaction between all of these disciplines was essential to the success of these microprocessors. The talk will discuss some of the key techniques and their evolution, which enabled Alpha microprocessors to deliver leadership performance.
Abstract: During the last few years, the complexity and size of digital systems has increased so drastically that guaranteeing correctness is no longer a trivial task. Formal verification provides a framework that allows to reason formally about the behavior of a system. The implementation is checked against a set of properties that have to be satisfied. Typically, a digital system is modeled as a collection of interacting subsystems. As new subsystems are added the number of states may grow exponentially. This is known as the "state domain explosion" problem. Abstract interpretation is one approach to alleviate the state domain explosion problem. The main idea is to interpret the behavior of a system in a different abstracted (and therefore simplified) system with fewer states. The challenges for these techniques are the range of applicability and the degree of automation provided to the designer. So far, automatic abstraction techniques have been applied in very specific examples, and usually require the user intervention to guide the process. This presentation proposes an automatic abstraction paradigm to verify generic digital systems, as well as a practical implementation.
Bio: Abelardo Pardo received his BS and MS degrees in Computer Science at the Polytechnic University of Catalonia, Barcelona, Spain, in 1991 and his Ph.D. at the University of Colorado at Boulder in 1997. His areas of interest include formal verification of reactive systems and algorithms for automatic abstraction for verification. He is currently working at the Boston Advanced Development Labs of Mentor Graphics in Billerica, MA.
Abstract: For some years it has been known that surface shape can be recovered from the dynamics of apparent contours (a.k.a. profiles, outlines, occluding contours) provided motion is known. I shall review this fact and go on to describe some recent work which suggests, rather tentatively, that motion can also be recovered from apparent contours. As yet only one very simple case has been proved to work, but experimental work by K. Astrom in Sweden suggests that in favourable conditions, the new algorithm does converge, and sometimes to the right answer! This is joint work with Roberto Cipolla in Cambridge (U.K.) and Astrom.
Back to the LEMS homepage.