Inferring perceptual saliency fields from viewpoint-dependent recognition data.

We present an algorithm for computing the relative perceptual saliencies of the features of a three-dimensional object using either goodness-of-view scores measured at several viewpoints or perceptual similarities among several object views. This technique addresses the inverse, ill-posed version of the direct problem of predicting goodness-of-view scores or viewpoint similarities when the object features are known. On the basis of a linear model for the direct problem, we solve the inverse problem using the method of regularization. The critical assumption we make to regularize the solution is that perceptual salience varies slowly on the surface of the object. The salient regions derived using this assumption empirically indicate what object structures are important in human three-dimensional object perception, a domain where theories typically have been based on somewhat ad hoc features.     

一种基于感知物体的场景分析注意机制

基于物体的选择性注意在心理学领域正日益为广大研究人员所认可,而计算机视觉领域中现有的注意模型大多数是基于特征的,或者是基于空间的.本文给出了一种基于物体的选择性注意计算模型.该模型将“感知物体”作为引起注意的基本单元,并给出了感知物体及其邻域的定义.该注意模型包括两个步骤：（1）在给定图像中选择第一个注视点;（2）在整幅图像中实现注视点的有效转移.在该注意模型中,感知物体与其邻域之间灰度值的绝对差异--对比度,被作为该感知物体显著性的一种度量,并且注视点在图像中的转移顺序是由每个感知物体的显著度的次序来决定的.该模型的优点有：首先,由于该模型是完全基于感知物体的,使得其输出结果可以很容易地应用到物体识别、图像分割和场景分析中;其次,该模型是多尺度的,也就是说,它可以根据实际任务的需要进行适当的调整.大量的真实图像实验表明,所提出的模型具有一定的合理性.     

Discovering hierarchical object models from captioned images

We address the problem of automatically learning the recurring associations between the visual structures in images and the words in their associated captions, yielding a set of named object models that can be used for subsequent image annotation. In previous work, we used language to drive the perceptual grouping of local features into configurations that capture small parts (patches) of an object. However, model scope was poor, leading to poor object localization during detection (annotation), and ambiguity was high when part detections were weak. We extend and significantly revise our previous framework by using language to drive the perceptual grouping of parts, each a configuration in the previous framework, into hierarchical configurations that offer greater spatial extent and flexibility. The resulting hierarchical multipart models remain scale, translation and rotation invariant, but are more reliable detectors and provide better localization. Moreover, unlike typical frameworks for learning object models, our approach requires no bounding boxes around the objects to be learned, can handle heavily cluttered training scenes, and is robust in the face of noisy captions, i.e., where objects in an image may not be named in the caption, and objects named in the caption may not appear in the image. We demonstrate improved precision and recall in annotation over the non-hierarchical technique and also show extended spatial coverage of detected objects.     

自适应尺度目标跟踪算法

针对复杂情况下变尺度目标跟踪问题,提出一种基于粒子滤波的自适应尺度目标跟踪算法.根据参考目标的颜色分布,将参考目标分为多个区域,每个区域的颜色分布用高斯模型表示,区域的位置关系构成了对参考目标的空间约束;根据目标分割区域的颜色分布和空间约束关系构造目标外观模型,结合粒子滤波搜索目标位置并检测目标的尺度变化.目标外观模型同时包含了空间及颜色信息,提高了跟踪算法在复杂情况下检测目标尺度变化的可靠性和准确性.实验结果表明,该算法在目标具有明显尺度变化、姿态改变和部分遮挡的情况下,可以获得准确和鲁棒的跟踪结果.     

Language and Space: a two-level semantic approach based on principles of ontological engineering

An increasing number of applications for dialogue systems presuppose an ability to deal appropriately with space. Dialogues with assistance systems, intelligent mobility devices and navigation systems all commonly involve the use of spatial language. For smooth interaction, this spatial language cannot be interpreted ‘in the abstract’—it must instead be related directly to a user’s physical location, orientation, goals and needs and be embedded appropriately in a system’s interaction. This is far from straightforward. The situated interpretation of natural language concerning space, spatial relationships and spatial activities represents an unsolved challenge at this time. Despite extensive work on spatial language involving many disciplines, there are no generally accepted accounts that provide support for the kind of flexible language use observed in real human-human spatial dialogues. In this paper, I review some recent approaches to the semantics for natural language expressions concerning space in order to motivate a two-level semantic-based approach to the interpretation of spatial language. This draws on a new combination of natural language processing and principles of ontological engineering and stands as a foundation for more sophisticated and natural dialogue system behavior where spatial information is involved.     

A computational model of perceptual saliency for 3D objects in virtual environments

When giving directions to the location of an object, people typically use other attractive objects as reference, that is, reference objects. With the aim to select proper reference objects, useful for locating a target object within a virtual environment (VE), a computational model to identify perceptual saliency is presented. Based on the object’s features with the major stimulus for the human visual system, three basic features of a 3D object (i.e., color, size, and shape) are individually evaluated and then combined to get a degree of saliency for each 3D object in a virtual scenario. An experiment was conducted to evaluate the extent to which the proposed measure of saliency matches with the people’s subjective perception of saliency; the results showed a good performance of this computational model.     

一种基于物体的视觉注意计算模型

为解决基于空间的视觉注意计算模型存在的注意目标不完整、容易转移到无意义区域等问题,提出一种结合空间显著性的基于物体的视觉注意计算模型。检测图像的边缘信息,根据空间视觉显著性度量结果,提取显著值高的封闭边缘,得到感知物体的轮廓。根据各感知物体的大小、位置和显著程度计算其注意度。注意焦点按照注意度递减的顺序在各感知物体之间进行转移。在多幅自然图像上进行实验验证,实验结果表明该模型具有和人类视觉特性相符合的注意效果。     

Static use-based object confinement

The confinement of object references is a significant security concern for modern programming languages. We define a language that serves as a uniform model for a variety of confined object reference systems. A use-based approach to confinement is adopted, which we argue is more expressive than previous communication-based approaches. We then develop a readable, expressive type system for static analysis of the language, along with a type safety result demonstrating that run-time checks can be eliminated. The language and type system thus serve as a reliable, declarative, and efficient foundation for secure capability-based programming and object confinement .     

Realm-based spatial data types: The ROSE algebra

Spatial data types or algebras for database systems should (1) be fully general, that is, closed under set operations, (2) have formally defined semantics, (3) be defined in terms of finite representations available in computers, (4) offer facilities to enforce geometric consistency of related spatial objects, and (5) be independent of a particular DBMS data model, but cooperate with any. We present an algebra that usesrealms as geometric domains underlying spatial data types. A realm, as a general database concept, is a finite, dynamic, user-defined structure underlying one or more system data types. Problems of numerical robustness and topological correctness are solved within and below the realm layer so that spatial algebras defined above a realm have very nice algebraic properties. Realms also interact with a DMBS to enforce geometric consistency on object creation or update. The ROSE algebra is defined on top of realms and offers general types to represent point, line, and region features, together with a comprehensive set of operations. It is described within a polymorphic type system and interacts with a DMBS data model and query language through an abstractobject model interface. An example integration of ROSE into the object-oriented data model O₂ and its query language is presented.     

Efficient Subwindow Search: A Branch and Bound Framework for Object Localization

Most successful object recognition systems rely on binary classification, deciding only if an object is present or not, but not providing information on the actual object location. To estimate the object's location, one can take a sliding window approach, but this strongly increases the computational cost because the classifier or similarity function has to be evaluated over a large set of candidate subwindows. In this paper, we propose a simple yet powerful branch and bound scheme that allows efficient maximization of a large class of quality functions over all possible subimages. It converges to a globally optimal solution typically in linear or even sublinear time, in contrast to the quadratic scaling of exhaustive or sliding window search. We show how our method is applicable to different object detection and image retrieval scenarios. The achieved speedup allows the use of classifiers for localization that formerly were considered too slow for this task, such as SVMs with a spatial pyramid kernel or nearest-neighbor classifiers based on the chi^2 distance. We demonstrate state-of-the-art localization performance of the resulting systems on the UIUC Cars data set, the PASCAL VOC 2006 data set, and in the PASCAL VOC 2007 competition.     

基于注意覆盖的感兴趣区域增强策略

提出一种基于注意覆盖的感兴趣区域增强策略。将感知表面作为注意选择的基本单元,将自顶向下的注意信号引入表面填充机制,利用非线性扩散机制在感知表面形成一种形状拟合分布,使得被注意的感知物体活性得到增强。仿真结果表明,该策略可以有效地增强感兴趣区域,具有神经生理和心理学合理性,输出结果可用于区域分割、目标识别和场景分析。     

A psychophysical approach to assessing the quality of antialiasedimages

The assumption that antialiasing destroys useful visual information about object features is challenged in three experiments that examine the effects of antialiasing on the visual information for object location and motion. The results show that proper antialiasing eliminates the spurious visual information produced by sampling processes in image synthesis and allows the viewer's visual system to produce a precise representation of object location and a continuous representation of object motion. This suggests that in designing imagery systems, simply increasing the spatial and temporal addressability and resolution beyond limits set by the human visual system will have a negligible impact on image quality, but that effective use of antialiasing techniques could allow visual information about object features to be presented with great fidelity     

Object-based visual attention for computer vision

In this paper, a novel model of object-based visual attention extending Duncan's Integrated Competition Hypothesis [Phil. Trans. R. Soc. London B 353 (1998) 1307-1317] is presented. In contrast to the attention mechanisms used in most previous machine vision systems which drive attention based on the spatial location hypothesis, the mechanisms which direct visual attention in our system are object-driven as well as feature-driven. The competition to gain visual attention occurs not only within an object but also between objects. For this purpose, two new mechanisms in the proposed model are described and analyzed in detail. The first mechanism computes the visual salience of objects and groupings; the second one implements the hierarchical selectivity of attentional shifts. The results of the new approach on synthetic and natural images are reported.     

An Expectation-Based Framework of Object Schemas and Port-Based Agents for Disparate Feedback Assimilation

Robotic manipulation systems that operate in unstructured environments must be responsive to feedback from sensors that are disparate in both location and modality. This paper describes a distributed framework for assimilating the disparate feedback provided by force and vision sensors, including active vision sensors, for robotic manipulation systems. The main components of the expectation-based framework include object schemas and port-based agents. Object schemas represent the manipulation task internally in terms of geometric models with attached sensor mappings. Object schemas are dynamically updated by sensor feedback, and thus provide an ability to perform three dimensional spatial reasoning during task execution. Because object schemas possess knowledge of sensor mappings, they are able to both select appropriate sensors and guide active sensors based on task characteristics. Port-based agents are the executors of reference inputs provided by object schemas and are defined in terms of encapsulated control strategies. Experimental results demonstrate the capabilities of the framework in two ways: the performance of manipulation tasks with active camera-lens systems, and the assimilation of force and vision sensory feedback.     

汉语中方位参考点恢复研究

方位参考点恢复是自然语言空间语义理解中十分重要问题 .方位参考点恢复是在篇章中找方位词的参考点并补充上,得到完整的空间表达式 .目前,自然语言处理技术大多面向句子级,导致省略参考点空间表达式独立出现,使空间语义理解困难 .方位参考点恢复无疑可以解决类似问题 .在此提出基于有限知识的方位参考点恢复方法 .在句法分析基础上,以知网为常识库,结合有限知识识别空间表达式以及恢复方位参考点 .实验结果表明该方法比较令人满意 .     

Spatial Plateau Algebra for implementing fuzzy spatial objects in databases and GIS: Spatial plateau data types and operations

Many geographical applications have to deal with spatial objects that reveal an intrinsically vague or fuzzy nature. A spatial object is fuzzy if locations exist that cannot be assigned completely to the object or to its complement. Spatial database systems and Geographical Information Systems (GIS) are currently unable to cope with this kind of data. Based on an available abstract data model of fuzzy spatial data types for fuzzy points, fuzzy lines, and fuzzy regions that leverages fuzzy set theory and fuzzy point set topology, this article proposes a Spatial Plateau Algebra that provides spatial plateau data types as an implementation of fuzzy spatial data types. Each spatial plateau object consists of a finite number of crisp counterparts that are all adjacent or disjoint to each other, are associated with different membership values, and hence form different plateaus. The formal framework and the implementation are based on well known, exact models and implementations of crisp spatial data types. Spatial plateau operations as geometric operations on spatial plateau objects are expressed as a combination of geometric operations on the underlying crisp spatial objects. This article offers a conceptually clean foundation for implementing a database extension for fuzzy spatial objects and their operations, and demonstrates the embedding of these new data types as attribute data types in a database schema as well as the incorporation of fuzzy spatial operations into a database query language.     

Put: language-based interactive manipulation of objects

Our approach to scene generation capitalizes the expressive power of natural language by separating its aptness in specifying spatial relations from the difficulties of understanding text. We are implementing an object-placement system called Put that uses a combination of linguistic commands and direct manipulation. The system is language-based, meaning that its design and structure are guided by natural language. Our approach (inspired by research in cognitive linguistics) is to analyze the natural use of spatial relations, define a well-understood class of fundamental relationships, and gradually build a coherent and natural spatial-manipulation system. Just a few simple spatial relationships, such as in, on, and at, parameterized by a limited number of environmental variables can provide comfortable object manipulation. These natural commands can be used to quickly prototype a complex scene and constrain object placement. We believe that we have an extensible, predictable, and computationally feasible environment for object manipulation. We have focused first on spatial relationships because they are fundamental to many conceptual domains beyond object placement, including motion and time. These particular domains are very important to areas of computer graphics such as animation. Uses of spatial relationships in these areas can be quite complex. We briefly introduce the complexities of understanding spatial relations and summarize related work. Then we describe the core of the Put placement system, followed by its linguistic, procedural, and interactive interfaces. We conclude by discussing future enhancements to the system     

Knowledge discovery from spatial transactions

We propose a general mechanism to represent the spatial transactions in a way that allows the use of the existing data mining methods. Our proposal allows the analyst to exploit the layered structure of geographical information systems in order to define the layers of interest and the relevant spatial relations among them. Given a reference object, it is possible to describe its neighborhood by considering the attribute of the object itself and the objects related by the chosen relations. The resulting spatial transactions may be either considered like “traditional” transactions, by considering only the qualitative spatial relations, or their spatial extension can be exploited during the data mining process. We explore both these cases. First we tackle the problem of classifying a spatial dataset, by taking into account the spatial component of the data to compute the statistical measure (i.e., the entropy) necessary to learn the model. Then, we consider the task of extracting spatial association rules, by focusing on the qualitative representation of the spatial relations. The feasibility of the process has been tested by implementing the proposed method on top of a GIS tool and by analyzing real world data.