基于稀疏表示的单帧运动图像盲复原

提出一种基于稀疏表示的单帧运动盲复原方法,它充分利用自然图像中存在的各种先验知识进行求解。该方法分为模糊核估计和图像修复两个阶段。在估计模糊核时,它运用shock滤波器从模糊图像中预测出清晰边缘,以此指导全局图像的复原,并运用多尺度策略来解决大模糊核问题。在图像修复阶段,运用稀疏表示理论对复原图像进行降噪和重建,最终提高图像复原质量。实验结果表明,在不同噪声和模糊核条件下,该算法能有效消除运动模糊。     

Monocular human motion tracking with discriminative sparse representation

In this work, we address the problem of monocular tracking the human motion based on the discriminative sparse representation. The proposed method jointly trains the dictionary and the discriminative linear classifier to separate the human being from the background. We show that using the online dictionary learning, the tracking algorithm can adapt the variation of human appearance and background environment. We compared the proposed method with four state-of-the-art tracking algorithms on eight benchmark video clips (Faceocc, Sylv, David, Singer, Girl, Ballet, OneLeaveShopReenter2cor, and ThreePastShop2cor). Qualitative and quantitative experimental validation results are discussed at length. The proposed algorithm for human tracking achieves superior tracking results, and a Matlab run time on a standard desktop machine of four frames per second.     

运动序列的时空结构特征表示模型

运动序列是一种与运动信号相关的多维时间序列,各个维度序列之间具有高耦合性的特点。现有的多维序列表征方法大多基于维度间相互独立的假设或缺乏可解释性,为此,提出一种适用于运动序列的时空结构特征表示模型及其两阶段构造方法。首先,基于空间变化事件的转换方法,将多维时间序列变换成一维事件序列,以保存序列中的空间结构特性。接着,定义了一种时空结构特征的无监督挖掘算法。基于新定义的表示度度量,该算法从事件序列中提取一组具有代表性的低冗余变长事件元组为时空结构特征。在多个人类行为识别数据集上的实验结果表明,与现有多维时间序列表示方法相比,新模型的特征集更具代表性,在运动序列模式识别领域可以有效提升分类精度。     

Affine shape representation from motion through reference points

This paper shows how an affine representation of spatial configuration is obtained from a pair of projection views. Calibration of cameras and knowledge of the camera's motion are not necessary; however, some preselected reference points and their correspondences are needed. Projective and affine geometry invariants are trickily manipulated to do the affine reconstruction. The method is thus geometrically constructive. When it is compared with the solution proposed in 1989 by J.J. Koenderink and A.J. Van Doorn (Affine Structure from Motion, Technical Report, Utrect University), the method provides a viewpoint-independent affine representation under parallel projections. Further, we investigate the central-projection case in which, with three additional special reference points, the same affine reconstruction can be done. We also discuss some important applications of this viewpoint independence of shape representation.     

Semantic representation and correspondence for state-based motion transition

Consistent transition algorithms preserve salient source motion features by establishing feature-based correspondence between motions and accordingly warping them before interpolation. These processes are commonly dubbed as preprocessing in motion transition literature. Current transition methods suffer from a lack of economical and generic preprocessing algorithms. Classical computer vision methods for human motion classification and correspondence are too computationally intensive for computer animation. The paper proposes an analytical framework that combines low-level kinematics analysis and high-level knowledge-based analysis to create states that provide coherent snapshots of body-parts active during the motion. These states are then corresponded via a globally optimal search tree algorithm. The framework proposed here is intuitive, controllable, and delivers results in near realtime. The validity and performance of the proposed system are tangibly proven with extensive experiments.     

表征空间中的机器人分层运动规划

由于对机器人的任务要求日趋复杂和多变,如何使机器人具备灵活的配置和运动规划能力,以适应复杂任务的需求,成为了目前运动规划领域所研究的核心问题.传统的基于任务空间和配置空间的建模方法虽然在机器人运动规划领域得到了非常广泛的应用,但在用于解决复杂规划任务时无法对不可行任务进行进一步地处理.本文在表征空间模型的基础上,提出了一种分层的运动规划算法,一方面借助于表征空间维度的扩展,使对运动规划任务的描述更为灵活;另一方面通过任务层与运动层的循环交互,使生成的路径满足更高层次和更丰富的任务要求.在仿人机器人和多机器人系统上的应用结果表明了本文所提算法的有效性.     

Browsing databases with constraint hierarchies

Electronic markets rely on database technology for efficient search for products, services, customers, and trading partners. Yet, database technology was not designed to build electronic markets, and it often fails to fully support market search activities. Market search often requires browsing and incremental search, attribute and constraint tradeoffs, approximate matches, and incorporation of user and task characteristics into searches. None of these requirements are supported directly by the database technology. Three major extensions to database technology are proposed to effectively support electronic markets. First, incremental search techniques are proposed to allow users to browse databases, and to move from solution to solution by following attribute and constraint tradeoffs. These techniques require systems to have some knowledge of similarity, distance, direction, and tradeoffs. Second, typing and stereotyping techniques are introduced to allow users to relate task and user characteristics to product attributes. These techniques require integration of typing systems with database searches. Third, a comprehensive search strategy is developed that combines incremental search and stereotyping techniques by utilizing product, task, and user ontologies.     

Offline reinforcement learning with task hierarchies

In this work, we build upon the observation that offline reinforcement learning (RL) is synergistic with task hierarchies that decompose large Markov decision processes (MDPs). Task hierarchies can allow more efficient sample collection from large MDPs, while offline algorithms can learn better policies than the so-called “recursively optimal” or even hierarchically optimal policies learned by standard hierarchical RL algorithms. To enable this synergy, we study sample collection strategies for offline RL that are consistent with a provided task hierarchy while still providing good exploration of the state-action space. We show that naïve extensions of uniform random sampling do not work well in this case and design a strategy that has provably good convergence properties. We also augment the initial set of samples using additional information from the task hierarchy, such as state abstraction. We use the augmented set of samples to learn a policy offline. Given a capable offline RL algorithm, this policy is then guaranteed to have a value greater than or equal to the value of the hierarchically optimal policy. We evaluate our approach on several domains and show that samples generated using a task hierarchy with a suitable strategy allow significantly more sample-efficient convergence than standard offline RL. Further, our approach also shows more sample-efficient convergence to policies with value greater than or equal to hierarchically optimal policies found through an online hierarchical RL approach.     

Reconfiguring system hierarchies with multi-valued logic

We propose a mathematical formalism to define a system hierarchy and choose a way to reconfigure it based on multi-valued logic. We consider an application of this formalism to control over a group surveying the danger zone after a natural disaster.     

Human motions representation and recognition by directional motion history images

In this paper, we describe a technique for representing and recognizing human motions using directional motion history images. A motion history image is a single human motion image produced by superposing binarized successive motion image frames so that older frames may have smaller weights. It has, however, difficulty that the latest motion overwrites older motions, resulting in inexact motion representation and therefore incorrect recognition. To overcome this difficulty, we propose directional motion history images which describe a motion with respect to four directions of movement, i.e. up, down, right and left, employing optical flow. The directional motion history images are thus a set of four motion history images defined on four optical flow images. Experimental results show that the proposed technique achieves better performance in the recognition of human motions than the existent motion history images. This work was presented in part at the 13th International Symposium on Artificial Life and Robotics, Oita, Japan, January 31–February 2, 2008     

Layered 4D representation and voting for grouping from motion

We address the problem of perceptual grouping from motion cues by formulating it as a motion layers inference from a sparse and noisy point set in a 4D space. Our approach is based on a layered 4D representation of data, and a voting scheme for token communication, within a tensor voting computational framework. Given two sparse sets of point tokens, the image position and potential velocity of each token are encoded into a 4D tensor. By enforcing the smoothness of motion through a voting process, the correct velocity is selected for each input point as the most salient token. An additional dense voting step allows for the inference of a dense representation in terms of pixel velocities, motion regions, and boundaries. Using a 4D space for this tensor voting approach is essential since it allows for a spatial separation of the points according to both their velocities and image coordinates. Unlike most other methods that optimize certain objective functions, our approach is noniterative and, therefore, does not suffer from local optima or poor convergence problems. We demonstrate our method with synthetic and real images, by analyzing several difficult cases-opaque and transparent motion, rigid and nonrigid motion, curves and surfaces in motion.     

A body model server for human motion capture and representation

This paper presents a body model server (BMS) that provides real-time access to the position and posture of a person's torso, arms, hands, head, and eyes. It can be accessed by clients over a network. The BMS is designed to function as a device-independent data-layer between the sensing devices and client applications that require real-time human motion data, such as animation control. It can provide clients with accurate information at up to 40 Hz. For data collection, the model uses four magnetic position/orientation sensors, two data-gloves, and an eye-tracker. The BMS combines the data-streams from the sensors and transforms them into snapshots of the user's upper-body pose. A geometric model made up of joints and segments structures the input. Posture of the body is represented by joint angles. Two unique characteristics of our approach are the use of the implicit, geometric constraints of the sensed body to simplify the computation of the unmeasured joint angles, and the use of time-stamped data that allow synchronization with other data streams, e.g., speech input. This paper describes the architecture of the BMS, including the management of multiple input devices, the representation and computation of the position and joint angle data, and the client-server interface.     

Constraint hierarchies

Constraints allow programmers and users to state declaratively a relation that should be maintained, rather than requiring them to write procedures to maintain the relation themselves. They are thus useful in such applications as programming languages, user interface toolkits, and simulation packages. In many situations, it is desirable to be able to state bothrequired andpreferential constraints. The required constraints must hold. Since the other constraints are merely preferences, the system should try to satisfy them if possible, but no error condition arises if it cannot. Aconstraint hierarchy consists of a set of constraints, each labeled as either required or preferred at some strength. An arbitrary number of different strengths is allowed. In the discussion of a theory of constraint hierarchies, we present alternate ways of selecting among competing possible solutions, and prove a number of propositions about the relations among these alternatives. We then outline algorithms for satisfying constraint hierarchies, and ways in which we have used constraint hierarchies in a number of programming languages and systems.     

Constraint hierarchies

Constraints allow programmers and users to state declaratively a relation that should be maintained, rather than requiring them to write procedures to maintain the relation themselves. They are thus useful in such applications as programming languages, user interface toolkits, and simulation packages. In many situations, it is desirable to be able to state bothrequired andpreferential constraints. The required constraints must hold. Since the other constraints are merely preferences, the system should try to satisfy them if possible, but no error condition arises if it cannot. Aconstraint hierarchy consists of a set of constraints, each labeled as either required or preferred at some strength. An arbitrary number of different strengths is allowed. In the discussion of a theory of constraint hierarchies, we present alternate ways of selecting among competing possible solutions, and prove a number of propositions about the relations among these alternatives. We then outline algorithms for satisfying constraint hierarchies, and ways in which we have used constraint hierarchies in a number of programming languages and systems.     

MOLAP带层次的维聚集实现

随着数据库技术的广泛应用和发展,产生了数据仓库、联机分析处理等一系列新技术,并且在实践中得以逐步应用.对于不同类型的联机分析处理技术的研究应用,以关系型的居多,而多维型的研究应用相对较少.通过对多维联机分析处理进行的研究可知,维聚集的实现是一个重点,而带层次的维聚集的实现是一个难点.探讨了多维联机分析处理带层次的维聚集的实现,在进行了分析的基础上,给出了类的设计,之后根据算法用代码予以实现,通过实例进行了验证.     

A relationship between difference hierarchies and relativized polynomial hierarchies

Chang and Kadin have shown that if the difference hierarchy over NP collapses to levelk, then the polynomial hierarchy (PH) is equal to thekth level of the difference hierarchy over ₂ ^p . We simplify their poof and obtain a slightly stronger conclusion: if the difference hierarchy over NP collapses to levelk, then PH collapses to (P _(k–1) ^NP )^NP, the class of sets recognized in polynomial time withk – 1 nonadaptive queries to a set in NP^NP and an unlimited number of queries to a set in NP. We also extend the result to classes other than NP: For any classC that has _m ^p -complete sets and is closed under _conj ^p -and _m ^NP -reductions (alternatively, closed under _disj ^p -and _m ^co-NP -reductions), if the difference hierarchy overC collapses to levelk, then PH ^C = (P _(k–1)–tt ^NP ) ^C . Then we show that the exact counting class C_P is closed under _disj ^p - and _m ^co-NP -reductions. Consequently, if the difference hierarchy over C_P collapses to levelk, then PH^PP(= PH^C_P) is equal to (P _(k–1)–tt ^NP )^PP. In contrast, the difference hierarchy over the closely related class PP is known to collapse.Finally we consider two ways of relativizing the bounded query class P _k–tt ^NP : the restricted relativization P _k–tt ^{NP C} and the full relativization (P _k–tt ^NP ) ^C . IfC is NP-hard, then we show that the two relativizations are different unless PH ^C collapses.Richard Beigel was supported in part by NSF Grants CCR-8808949 and CCR-8958528. Richard Chang was supported in part by NSF Research Grant CCR 88-23053. This work was done while Mitsunori Ogiwara was at the Department of Information Science, Tokyo Institute of Technology, Tokyo, Japan.